Classifiers_NSL_KDD.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Classifiers_NSL-KDD.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "7MblIbMu31gO"
      },
      "source": [
        "from sklearn.metrics import accuracy_score # for calculating accuracy of model\n",
        "from sklearn.model_selection import train_test_split # for splitting the dataset for training and testing\n",
        "from sklearn.metrics import classification_report # for generating a classification report of model\n",
        "import pickle # saving and loading trained model\n",
        "from os import path\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DriCvhiI4JTH"
      },
      "source": [
        "bin_data = pd.read_csv('datasets/bin_data.csv')\n",
        "bin_data.drop(bin_data.columns[0],axis=1,inplace=True)\n",
        "multi_data = pd.read_csv('datasets/multi_data.csv')\n",
        "multi_data.drop(multi_data.columns[0],axis=1,inplace=True)\n",
        "le1_classes_ = np.load('labels/le1_classes.npy',allow_pickle=True)\n",
        "le2_classes_ = np.load('le2_classes.npy',allow_pickle=True)"
      ],
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "X6gO0-O_ev95"
      },
      "source": [
        "# **Linear Support Vector Machine Classifier (Binary Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "i46cQs7J9sR8"
      },
      "source": [
        "# importing library for support vector machine classifier\n",
        "from sklearn.svm import SVC"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G73DgDbpfHdP"
      },
      "source": [
        "X = bin_data.iloc[:,0:93].to_numpy() # dataset excluding target attribute (encoded, one-hot-encoded,original)\n",
        "Y = bin_data['intrusion'] # target attribute"
      ],
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-8CHud3rbmEB"
      },
      "source": [
        "# splitting the dataset 75% for training and 25% testing\n",
        "X_train, X_test, y_train, y_test = train_test_split(X,Y, test_size=0.25, random_state=42) "
      ],
      "execution_count": 24,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OOg12Nn7e8do",
        "outputId": "e35d9091-3abb-4a6c-8225-229046890d66"
      },
      "source": [
        "# using kernel as linear \n",
        "lsvm = SVC(kernel='linear',gamma='auto') \n",
        "lsvm.fit(X_train,y_train) # training model on training dataset"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "SVC(C=1.0, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,\n",
              "    decision_function_shape='ovr', degree=3, gamma='auto', kernel='linear',\n",
              "    max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
              "    tol=0.001, verbose=False)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 116
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "40emxT1YEW1B",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "51238a89-2564-4341-c04e-5078ce99976b"
      },
      "source": [
        "pkl_filename = \"./models/lsvm_binary.pkl\"\n",
        "if (not path.isfile(pkl_filename)):\n",
        "  # saving the trained model to disk\n",
        "  with open(pkl_filename, 'wb') as file:\n",
        "    pickle.dump(lsvm, file)\n",
        "  print(\"Saved model to disk\")\n",
        "  # loading the trained model from disk\n",
        "  with open(pkl_filename, 'rb') as file:\n",
        "    lsvm = pickle.load(file)\n",
        "  print(\"Loaded model from disk\") "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "NmtWOt9ZipS2",
        "outputId": "f28290f8-7937-487b-c057-85147e4e91a5"
      },
      "source": [
        "y_pred = lsvm.predict(X_test) # predicting target attribute on testing dataset\n",
        "ac = accuracy_score(y_test, y_pred)*100 # calculating accuracy of predicted data\n",
        "print(\"LSVM-Classifier Binary Set-Accuracy is \", ac)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "LSVM-Classifier Binary Set-Accuracy is  96.69778370483266\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "T_-KRKlrg8gS",
        "outputId": "aa6ae973-f345-4a01-d68f-942b102d07ba"
      },
      "source": [
        "# classification report\n",
        "print(classification_report(y_test, y_pred,target_names=le1_classes_))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "    abnormal       0.97      0.96      0.96     14720\n",
            "      normal       0.96      0.97      0.97     16774\n",
            "\n",
            "    accuracy                           0.97     31494\n",
            "   macro avg       0.97      0.97      0.97     31494\n",
            "weighted avg       0.97      0.97      0.97     31494\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FD827OpHPVkl"
      },
      "source": [
        "# **Quadratic Support Vector Machine Classifier (Binary Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jvIqsl-8PSBw",
        "outputId": "0d3bcea4-5eb2-4ee7-c1bf-4efe29ca8a0b"
      },
      "source": [
        "qsvm=SVC(kernel='poly',gamma='auto') # using kernal as polynomial for quadratic svm\n",
        "qsvm.fit(X_train,y_train) # training model on training dataset"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "SVC(C=1.0, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,\n",
              "    decision_function_shape='ovr', degree=3, gamma='auto', kernel='poly',\n",
              "    max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
              "    tol=0.001, verbose=False)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 120
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fx6kRBvlE-RI",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f3dbecc9-17fd-42b2-a51e-be0860d3a436"
      },
      "source": [
        "pkl_filename = \"./models/qsvm_binary.pkl\"\n",
        "if (not path.isfile(pkl_filename)):\n",
        "  # saving the trained model to disk \n",
        "  with open(pkl_filename, 'wb') as file:\n",
        "    pickle.dump(qsvm, file)\n",
        "  print(\"Saved model to disk\")\n",
        "  # loading the trained model from disk\n",
        "  with open(pkl_filename, 'rb') as file:\n",
        "    qsvm = pickle.load(file)\n",
        "  print(\"Loaded model from disk\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "i8uUCqtAPAKw",
        "outputId": "895da3ff-7c5f-4576-c832-6c18ba21cb14"
      },
      "source": [
        "y_pred=qsvm.predict(X_test) # predicting target attribute on testing dataset\n",
        "ac=accuracy_score(y_test, y_pred)*100 # calculating accuracy of predicted data\n",
        "print(\"QSVM-Classifier Binary Set-Accuracy is \", ac)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "QSVM-Classifier Binary Set-Accuracy is  95.71029402425859\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zfDbF9XNg6Bp",
        "outputId": "06717cbb-c99f-4757-e80c-7f3761b1274c"
      },
      "source": [
        "# classification report\n",
        "print(classification_report(y_test, y_pred,target_names=le1_classes_))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "    abnormal       0.99      0.92      0.95     14720\n",
            "      normal       0.93      0.99      0.96     16774\n",
            "\n",
            "    accuracy                           0.96     31494\n",
            "   macro avg       0.96      0.95      0.96     31494\n",
            "weighted avg       0.96      0.96      0.96     31494\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7bnFgn5XB_mD"
      },
      "source": [
        "# **K-nearest-neighbor Classifier (Binary Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CmdUXiaiB8K9"
      },
      "source": [
        "# importing library for K-neares-neighbor classifier\n",
        "from sklearn.neighbors import KNeighborsClassifier"
      ],
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AlJRTC_MCjoZ",
        "outputId": "ab206f5b-99f5-44ee-811f-c20ea715cde7"
      },
      "source": [
        "knn=KNeighborsClassifier(n_neighbors=5) # creating model for 5 neighbors\n",
        "knn.fit(X_train,y_train) # training model on training dataset"
      ],
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',\n",
              "                     metric_params=None, n_jobs=None, n_neighbors=5, p=2,\n",
              "                     weights='uniform')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 26
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "b7sDNWHBFG49",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b2b57879-1ca9-4991-c669-f6dcbced70a2"
      },
      "source": [
        "pkl_filename = \"./models/knn_binary.pkl\"\n",
        "if (not path.isfile(pkl_filename)):\n",
        "  # saving the trained model to disk\n",
        "  with open(pkl_filename, 'wb') as file:\n",
        "    pickle.dump(knn, file)\n",
        "  print(\"Saved model to disk\")\n",
        "  # loading the trained model from disk\n",
        "  with open(pkl_filename, 'rb') as file:\n",
        "    knn = pickle.load(file)\n",
        "  print(\"Loaded model from disk\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "VhDTePIsBm-t",
        "outputId": "036555c6-90a4-4cab-fada-9c55491864b7"
      },
      "source": [
        "y_pred=knn.predict(X_test) # predicting target attribute on testing dataset\n",
        "ac=accuracy_score(y_test, y_pred)*100 # calculating accuracy of predicted data\n",
        "print(\"KNN-Classifier Binary Set-Accuracy is \", ac)"
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "KNN-Classifier Binary Set-Accuracy is  98.54892995491204\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cSzvTXikg2pA",
        "outputId": "d3d01b41-937f-459f-8f9b-6a0c7e99d63b"
      },
      "source": [
        "# classification report\n",
        "print(classification_report(y_test, y_pred,target_names=le1_classes_))"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "    abnormal       0.99      0.98      0.98     14720\n",
            "      normal       0.99      0.99      0.99     16774\n",
            "\n",
            "    accuracy                           0.99     31494\n",
            "   macro avg       0.99      0.99      0.99     31494\n",
            "weighted avg       0.99      0.99      0.99     31494\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0mfnQh0jtn8a"
      },
      "source": [
        "# **Linear Discriminant Analysis Classifier (Binary Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3YtnOUXatx4V"
      },
      "source": [
        "# importing library for Linear Discriminant Analysis Model\n",
        "from sklearn.discriminant_analysis import LinearDiscriminantAnalysis"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "11-UNM_atzpz",
        "outputId": "3d301502-7b07-4832-a9d2-fc3db24d22d7"
      },
      "source": [
        "lda = LinearDiscriminantAnalysis() \n",
        "lda.fit(X_train, y_train)  # training model on training dataset"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "LinearDiscriminantAnalysis(n_components=None, priors=None, shrinkage=None,\n",
              "                           solver='svd', store_covariance=False, tol=0.0001)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 130
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dRY09jVFEJRy",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1d17764b-b692-438e-a131-e43981045ee9"
      },
      "source": [
        "pkl_filename = \"./models/lda_binary.pkl\"\n",
        "if (not path.isfile(pkl_filename)):\n",
        "  # saving trained model to disk\n",
        "  with open(pkl_filename, 'wb') as file:\n",
        "    pickle.dump(lda, file)\n",
        "  print(\"Saved model to disk\")\n",
        "  # loading trained model from disk\n",
        "  with open(pkl_filename, 'rb') as file:\n",
        "    lda = pickle.load(file)\n",
        "  print(\"Loaded model from disk\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hP0SNgdytqCA",
        "outputId": "8af55a76-ef24-4655-fae8-fbc3dea284fe"
      },
      "source": [
        "y_pred = lda.predict(X_test) # predicting target attribute on testing dataset\n",
        "ac=accuracy_score(y_test, y_pred)*100 # calculating accuracy of predicted data\n",
        "print(\"LDA-Classifier Set-Accuracy is \", ac)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "LDA-Classifier Set-Accuracy is  96.70730932876104\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "mYnAzR9xuJ7Q",
        "outputId": "759e4e5d-c600-4fe8-8179-ba8bade8bed6"
      },
      "source": [
        "# classification report\n",
        "print(classification_report(y_test, y_pred,target_names=le1_classes_))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "    abnormal       0.97      0.96      0.96     14720\n",
            "      normal       0.96      0.98      0.97     16774\n",
            "\n",
            "    accuracy                           0.97     31494\n",
            "   macro avg       0.97      0.97      0.97     31494\n",
            "weighted avg       0.97      0.97      0.97     31494\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fOETMr33uwN0"
      },
      "source": [
        "# **Quadratic Discriminant Analysis Classifier (Binary Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7kmCNlw_uvoL"
      },
      "source": [
        "# importing library for Quadratic Discriminant Analysis Model\n",
        "from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ALsaiYPYu9uV",
        "outputId": "d934a17d-d439-4c15-fb15-20a780202d35"
      },
      "source": [
        "qda = QuadraticDiscriminantAnalysis()\n",
        "qda.fit(X_train, y_train) # training model on training dataset"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/discriminant_analysis.py:691: UserWarning: Variables are collinear\n",
            "  warnings.warn(\"Variables are collinear\")\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "QuadraticDiscriminantAnalysis(priors=None, reg_param=0.0,\n",
              "                              store_covariance=False, tol=0.0001)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 135
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qeJNXcxjFrjS",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ae8250e5-b4db-4aa7-a3af-7bf472ab6b39"
      },
      "source": [
        "pkl_filename = \"./models/qda_binary.pkl\"\n",
        "if (not path.isfile(pkl_filename)):\n",
        "  # saving trained model to disk\n",
        "  with open(pkl_filename, 'wb') as file:\n",
        "    pickle.dump(qda, file)\n",
        "  print(\"Saved model to disk\")\n",
        "  # loading trained model from disk\n",
        "  with open(pkl_filename, 'rb') as file:\n",
        "    qda = pickle.load(file)\n",
        "  print(\"Loaded model from disk\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "NJ2ja7JpvFUg",
        "outputId": "9f58e21d-939e-4a01-de0c-54dad8513022"
      },
      "source": [
        "y_pred = qda.predict(X_test) # predicting target attribute on testing dataset\n",
        "ac=accuracy_score(y_test, y_pred)*100  # calculating accuracy of predicted data\n",
        "print(\"QDA-Classifier Set-Accuracy is \", ac)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "QDA-Classifier Set-Accuracy is  68.79088080269258\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0XYW6VI-vIQn",
        "outputId": "19ecb554-26f4-4178-ffd8-07010fed6a73"
      },
      "source": [
        "# classification report\n",
        "print(classification_report(y_test, y_pred,target_names=le1_classes_))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "    abnormal       1.00      0.33      0.50     14720\n",
            "      normal       0.63      1.00      0.77     16774\n",
            "\n",
            "    accuracy                           0.69     31494\n",
            "   macro avg       0.81      0.67      0.64     31494\n",
            "weighted avg       0.80      0.69      0.65     31494\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UcAd_oyaD4C6"
      },
      "source": [
        "# **Multi Layer Perceptron Classifier (Binary Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MXZbJB9HESF4"
      },
      "source": [
        "from keras.layers import Dense # importing dense layer\n",
        "from keras.models import Sequential #importing Sequential layer\n",
        "from keras.models import model_from_json # saving and loading trained model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Imkkjtu_Fnzr"
      },
      "source": [
        "X = bin_data.iloc[:,0:93].values # dataset excluding target attribute (encoded, one-hot-encoded,original)\n",
        "Y = bin_data[['intrusion']].values # target attribute"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gNx0tOV7FwfO"
      },
      "source": [
        "# splitting the dataset 75% for training and 25% testing\n",
        "X_train, X_test, y_train, y_test = train_test_split(X,Y, test_size=0.25, random_state=42)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9-5ChLnLEVqn"
      },
      "source": [
        "mlp = Sequential() # creating model\n",
        "\n",
        "# adding input layer and first layer with 50 neurons\n",
        "mlp.add(Dense(units=50, input_dim=X_train.shape[1], activation='relu'))\n",
        "# output layer with sigmoid activation\n",
        "mlp.add(Dense(units=1,activation='sigmoid'))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9ttL9NnuEsvL"
      },
      "source": [
        "# defining loss function, optimizer, metrics and then compiling model\n",
        "mlp.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GBlm3w0EzM24",
        "outputId": "1cce1b08-2d2a-4b13-d5f1-eecf9733a618"
      },
      "source": [
        "# summary of model layers\n",
        "mlp.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense_2 (Dense)              (None, 50)                4700      \n",
            "_________________________________________________________________\n",
            "dense_3 (Dense)              (None, 1)                 51        \n",
            "=================================================================\n",
            "Total params: 4,751\n",
            "Trainable params: 4,751\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TeIHduwLEzec",
        "outputId": "5dcbc619-340b-42af-9357-6d6ad88f930f"
      },
      "source": [
        "# training the model on training dataset\n",
        "history = mlp.fit(X_train, y_train, epochs=100, batch_size=5000,validation_split=0.2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "16/16 [==============================] - 1s 19ms/step - loss: 0.6427 - accuracy: 0.6652 - val_loss: 0.4590 - val_accuracy: 0.8974\n",
            "Epoch 2/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.4233 - accuracy: 0.8973 - val_loss: 0.3334 - val_accuracy: 0.9024\n",
            "Epoch 3/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.3165 - accuracy: 0.9029 - val_loss: 0.2670 - val_accuracy: 0.9091\n",
            "Epoch 4/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.2555 - accuracy: 0.9107 - val_loss: 0.2280 - val_accuracy: 0.9184\n",
            "Epoch 5/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.2195 - accuracy: 0.9179 - val_loss: 0.2013 - val_accuracy: 0.9245\n",
            "Epoch 6/100\n",
            "16/16 [==============================] - 0s 11ms/step - loss: 0.1969 - accuracy: 0.9232 - val_loss: 0.1809 - val_accuracy: 0.9308\n",
            "Epoch 7/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.1743 - accuracy: 0.9358 - val_loss: 0.1649 - val_accuracy: 0.9518\n",
            "Epoch 8/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.1603 - accuracy: 0.9523 - val_loss: 0.1524 - val_accuracy: 0.9589\n",
            "Epoch 9/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.1484 - accuracy: 0.9598 - val_loss: 0.1425 - val_accuracy: 0.9636\n",
            "Epoch 10/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.1403 - accuracy: 0.9631 - val_loss: 0.1346 - val_accuracy: 0.9669\n",
            "Epoch 11/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.1302 - accuracy: 0.9677 - val_loss: 0.1281 - val_accuracy: 0.9686\n",
            "Epoch 12/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.1235 - accuracy: 0.9691 - val_loss: 0.1229 - val_accuracy: 0.9687\n",
            "Epoch 13/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.1190 - accuracy: 0.9691 - val_loss: 0.1185 - val_accuracy: 0.9693\n",
            "Epoch 14/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.1150 - accuracy: 0.9692 - val_loss: 0.1147 - val_accuracy: 0.9695\n",
            "Epoch 15/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.1110 - accuracy: 0.9701 - val_loss: 0.1114 - val_accuracy: 0.9696\n",
            "Epoch 16/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.1083 - accuracy: 0.9700 - val_loss: 0.1084 - val_accuracy: 0.9697\n",
            "Epoch 17/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.1073 - accuracy: 0.9693 - val_loss: 0.1059 - val_accuracy: 0.9700\n",
            "Epoch 18/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.1035 - accuracy: 0.9703 - val_loss: 0.1035 - val_accuracy: 0.9702\n",
            "Epoch 19/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0985 - accuracy: 0.9716 - val_loss: 0.1014 - val_accuracy: 0.9702\n",
            "Epoch 20/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0997 - accuracy: 0.9700 - val_loss: 0.0994 - val_accuracy: 0.9703\n",
            "Epoch 21/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0968 - accuracy: 0.9710 - val_loss: 0.0975 - val_accuracy: 0.9702\n",
            "Epoch 22/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.0960 - accuracy: 0.9703 - val_loss: 0.0958 - val_accuracy: 0.9702\n",
            "Epoch 23/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0931 - accuracy: 0.9711 - val_loss: 0.0943 - val_accuracy: 0.9702\n",
            "Epoch 24/100\n",
            "16/16 [==============================] - 0s 8ms/step - loss: 0.0910 - accuracy: 0.9712 - val_loss: 0.0928 - val_accuracy: 0.9704\n",
            "Epoch 25/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0886 - accuracy: 0.9720 - val_loss: 0.0914 - val_accuracy: 0.9706\n",
            "Epoch 26/100\n",
            "16/16 [==============================] - 0s 8ms/step - loss: 0.0895 - accuracy: 0.9710 - val_loss: 0.0901 - val_accuracy: 0.9704\n",
            "Epoch 27/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0882 - accuracy: 0.9713 - val_loss: 0.0889 - val_accuracy: 0.9706\n",
            "Epoch 28/100\n",
            "16/16 [==============================] - 0s 8ms/step - loss: 0.0847 - accuracy: 0.9723 - val_loss: 0.0878 - val_accuracy: 0.9706\n",
            "Epoch 29/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0849 - accuracy: 0.9714 - val_loss: 0.0868 - val_accuracy: 0.9706\n",
            "Epoch 30/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0854 - accuracy: 0.9714 - val_loss: 0.0858 - val_accuracy: 0.9709\n",
            "Epoch 31/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0834 - accuracy: 0.9721 - val_loss: 0.0849 - val_accuracy: 0.9709\n",
            "Epoch 32/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.0809 - accuracy: 0.9726 - val_loss: 0.0840 - val_accuracy: 0.9706\n",
            "Epoch 33/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0814 - accuracy: 0.9723 - val_loss: 0.0832 - val_accuracy: 0.9706\n",
            "Epoch 34/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0806 - accuracy: 0.9728 - val_loss: 0.0825 - val_accuracy: 0.9707\n",
            "Epoch 35/100\n",
            "16/16 [==============================] - 0s 8ms/step - loss: 0.0821 - accuracy: 0.9716 - val_loss: 0.0819 - val_accuracy: 0.9705\n",
            "Epoch 36/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0785 - accuracy: 0.9728 - val_loss: 0.0812 - val_accuracy: 0.9705\n",
            "Epoch 37/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0798 - accuracy: 0.9719 - val_loss: 0.0805 - val_accuracy: 0.9705\n",
            "Epoch 38/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0774 - accuracy: 0.9725 - val_loss: 0.0800 - val_accuracy: 0.9705\n",
            "Epoch 39/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.0760 - accuracy: 0.9733 - val_loss: 0.0794 - val_accuracy: 0.9705\n",
            "Epoch 40/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0773 - accuracy: 0.9727 - val_loss: 0.0789 - val_accuracy: 0.9705\n",
            "Epoch 41/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0756 - accuracy: 0.9728 - val_loss: 0.0784 - val_accuracy: 0.9706\n",
            "Epoch 42/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0778 - accuracy: 0.9721 - val_loss: 0.0780 - val_accuracy: 0.9707\n",
            "Epoch 43/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0741 - accuracy: 0.9734 - val_loss: 0.0775 - val_accuracy: 0.9706\n",
            "Epoch 44/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0761 - accuracy: 0.9727 - val_loss: 0.0772 - val_accuracy: 0.9713\n",
            "Epoch 45/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.0734 - accuracy: 0.9740 - val_loss: 0.0767 - val_accuracy: 0.9716\n",
            "Epoch 46/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0744 - accuracy: 0.9738 - val_loss: 0.0763 - val_accuracy: 0.9719\n",
            "Epoch 47/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0733 - accuracy: 0.9739 - val_loss: 0.0760 - val_accuracy: 0.9723\n",
            "Epoch 48/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0740 - accuracy: 0.9743 - val_loss: 0.0756 - val_accuracy: 0.9727\n",
            "Epoch 49/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0738 - accuracy: 0.9740 - val_loss: 0.0753 - val_accuracy: 0.9727\n",
            "Epoch 50/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0720 - accuracy: 0.9749 - val_loss: 0.0750 - val_accuracy: 0.9727\n",
            "Epoch 51/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0716 - accuracy: 0.9749 - val_loss: 0.0747 - val_accuracy: 0.9731\n",
            "Epoch 52/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0721 - accuracy: 0.9749 - val_loss: 0.0744 - val_accuracy: 0.9732\n",
            "Epoch 53/100\n",
            "16/16 [==============================] - 0s 8ms/step - loss: 0.0729 - accuracy: 0.9747 - val_loss: 0.0741 - val_accuracy: 0.9733\n",
            "Epoch 54/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0706 - accuracy: 0.9756 - val_loss: 0.0738 - val_accuracy: 0.9734\n",
            "Epoch 55/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.0712 - accuracy: 0.9754 - val_loss: 0.0735 - val_accuracy: 0.9739\n",
            "Epoch 56/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0716 - accuracy: 0.9754 - val_loss: 0.0733 - val_accuracy: 0.9740\n",
            "Epoch 57/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0708 - accuracy: 0.9760 - val_loss: 0.0730 - val_accuracy: 0.9740\n",
            "Epoch 58/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0703 - accuracy: 0.9761 - val_loss: 0.0727 - val_accuracy: 0.9741\n",
            "Epoch 59/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.0691 - accuracy: 0.9766 - val_loss: 0.0724 - val_accuracy: 0.9741\n",
            "Epoch 60/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0703 - accuracy: 0.9765 - val_loss: 0.0723 - val_accuracy: 0.9742\n",
            "Epoch 61/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0703 - accuracy: 0.9760 - val_loss: 0.0720 - val_accuracy: 0.9745\n",
            "Epoch 62/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0689 - accuracy: 0.9768 - val_loss: 0.0717 - val_accuracy: 0.9746\n",
            "Epoch 63/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.0699 - accuracy: 0.9770 - val_loss: 0.0715 - val_accuracy: 0.9749\n",
            "Epoch 64/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0696 - accuracy: 0.9767 - val_loss: 0.0713 - val_accuracy: 0.9749\n",
            "Epoch 65/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0697 - accuracy: 0.9764 - val_loss: 0.0712 - val_accuracy: 0.9749\n",
            "Epoch 66/100\n",
            "16/16 [==============================] - 0s 11ms/step - loss: 0.0696 - accuracy: 0.9767 - val_loss: 0.0710 - val_accuracy: 0.9749\n",
            "Epoch 67/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0688 - accuracy: 0.9773 - val_loss: 0.0708 - val_accuracy: 0.9754\n",
            "Epoch 68/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0697 - accuracy: 0.9768 - val_loss: 0.0706 - val_accuracy: 0.9752\n",
            "Epoch 69/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0688 - accuracy: 0.9772 - val_loss: 0.0705 - val_accuracy: 0.9753\n",
            "Epoch 70/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0683 - accuracy: 0.9767 - val_loss: 0.0703 - val_accuracy: 0.9754\n",
            "Epoch 71/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0673 - accuracy: 0.9775 - val_loss: 0.0702 - val_accuracy: 0.9753\n",
            "Epoch 72/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0677 - accuracy: 0.9774 - val_loss: 0.0700 - val_accuracy: 0.9753\n",
            "Epoch 73/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0677 - accuracy: 0.9775 - val_loss: 0.0698 - val_accuracy: 0.9754\n",
            "Epoch 74/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0667 - accuracy: 0.9774 - val_loss: 0.0696 - val_accuracy: 0.9756\n",
            "Epoch 75/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0669 - accuracy: 0.9774 - val_loss: 0.0695 - val_accuracy: 0.9756\n",
            "Epoch 76/100\n",
            "16/16 [==============================] - 0s 11ms/step - loss: 0.0678 - accuracy: 0.9770 - val_loss: 0.0694 - val_accuracy: 0.9754\n",
            "Epoch 77/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.0675 - accuracy: 0.9775 - val_loss: 0.0692 - val_accuracy: 0.9757\n",
            "Epoch 78/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0654 - accuracy: 0.9787 - val_loss: 0.0691 - val_accuracy: 0.9759\n",
            "Epoch 79/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0657 - accuracy: 0.9776 - val_loss: 0.0690 - val_accuracy: 0.9756\n",
            "Epoch 80/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0670 - accuracy: 0.9774 - val_loss: 0.0688 - val_accuracy: 0.9758\n",
            "Epoch 81/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0665 - accuracy: 0.9782 - val_loss: 0.0687 - val_accuracy: 0.9757\n",
            "Epoch 82/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.0686 - accuracy: 0.9768 - val_loss: 0.0687 - val_accuracy: 0.9759\n",
            "Epoch 83/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.0641 - accuracy: 0.9785 - val_loss: 0.0686 - val_accuracy: 0.9758\n",
            "Epoch 84/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0656 - accuracy: 0.9783 - val_loss: 0.0684 - val_accuracy: 0.9758\n",
            "Epoch 85/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0669 - accuracy: 0.9777 - val_loss: 0.0683 - val_accuracy: 0.9760\n",
            "Epoch 86/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0643 - accuracy: 0.9784 - val_loss: 0.0682 - val_accuracy: 0.9758\n",
            "Epoch 87/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0666 - accuracy: 0.9775 - val_loss: 0.0681 - val_accuracy: 0.9759\n",
            "Epoch 88/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0664 - accuracy: 0.9781 - val_loss: 0.0680 - val_accuracy: 0.9760\n",
            "Epoch 89/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.0651 - accuracy: 0.9780 - val_loss: 0.0678 - val_accuracy: 0.9759\n",
            "Epoch 90/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0651 - accuracy: 0.9785 - val_loss: 0.0677 - val_accuracy: 0.9758\n",
            "Epoch 91/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0654 - accuracy: 0.9779 - val_loss: 0.0676 - val_accuracy: 0.9759\n",
            "Epoch 92/100\n",
            "16/16 [==============================] - 0s 11ms/step - loss: 0.0664 - accuracy: 0.9776 - val_loss: 0.0675 - val_accuracy: 0.9758\n",
            "Epoch 93/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0641 - accuracy: 0.9789 - val_loss: 0.0674 - val_accuracy: 0.9758\n",
            "Epoch 94/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0654 - accuracy: 0.9784 - val_loss: 0.0673 - val_accuracy: 0.9758\n",
            "Epoch 95/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0653 - accuracy: 0.9783 - val_loss: 0.0672 - val_accuracy: 0.9759\n",
            "Epoch 96/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0640 - accuracy: 0.9787 - val_loss: 0.0671 - val_accuracy: 0.9759\n",
            "Epoch 97/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0635 - accuracy: 0.9786 - val_loss: 0.0670 - val_accuracy: 0.9759\n",
            "Epoch 98/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0622 - accuracy: 0.9788 - val_loss: 0.0670 - val_accuracy: 0.9760\n",
            "Epoch 99/100\n",
            "16/16 [==============================] - 0s 9ms/step - loss: 0.0646 - accuracy: 0.9782 - val_loss: 0.0668 - val_accuracy: 0.9762\n",
            "Epoch 100/100\n",
            "16/16 [==============================] - 0s 10ms/step - loss: 0.0638 - accuracy: 0.9785 - val_loss: 0.0667 - val_accuracy: 0.9761\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Nn4cYSDKx1-o",
        "outputId": "dfdbefc7-758e-4837-9371-ad18e8408a18"
      },
      "source": [
        "filepath = './models/mlp_binary.json'\n",
        "weightspath = './weights/mlp_binary.h5'\n",
        "if (not path.isfile(filepath)):\n",
        "  # serialize model to JSON\n",
        "  mlp_json = mlp.to_json()\n",
        "  with open(filepath, \"w\") as json_file:\n",
        "    json_file.write(mlp_json)\n",
        "\n",
        "  # serialize weights to HDF5\n",
        "  mlp.save_weights(weightspath)\n",
        "  print(\"Saved model to disk\")\n",
        "\n",
        "  # load json and create model\n",
        "  json_file = open(filepath, 'r')\n",
        "  loaded_model_json = json_file.read()\n",
        "  json_file.close()\n",
        "  mlp = model_from_json(loaded_model_json)\n",
        "\n",
        "  # load weights into new model\n",
        "  mlp.load_weights(weightspath)\n",
        "  print(\"Loaded model from disk\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HRCboXxT04NZ"
      },
      "source": [
        "# defining loss function, optimizer, metrics and then compiling model\n",
        "mlp.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XYkgc0V6E2Cy",
        "outputId": "7b2d37c6-95c3-45e4-fc8e-9ff0ed63faea"
      },
      "source": [
        "# predicting target attribute on testing dataset\n",
        "test_results = mlp.evaluate(X_test, y_test, verbose=1)\n",
        "print(f'Test results - Loss: {test_results[0]} - Accuracy: {test_results[1]*100}')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "985/985 [==============================] - 1s 950us/step - loss: 0.0682 - accuracy: 0.9771\n",
            "Test results - Loss: 0.06512267887592316 - Accuracy: 97.79958128929138\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "8t6-ytIJE7th",
        "outputId": "9a8c68d0-8e6f-4a6c-c2a7-381cfcad4f85"
      },
      "source": [
        "# Plot of accuracy vs epoch of train and test dataset\n",
        "plt.plot(history.history['accuracy'])\n",
        "plt.plot(history.history['val_accuracy'])\n",
        "plt.title(\"Plot of accuracy vs epoch for train and test dataset\")\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='best')\n",
        "plt.savefig('plots/mlp_binary_accuracy.png')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "hun0u-pxE_q1",
        "outputId": "7d09e9d0-d89c-4413-bda3-1c37d71cb83f"
      },
      "source": [
        "# Plot of loss vs epoch of train and test dataset\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title('Plot of loss vs epoch for train and test dataset')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='best')\n",
        "plt.savefig('plots/mlp_binary_loss.png')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 312
        },
        "id": "BFiX0p1dy4BB",
        "outputId": "96fe5d29-dc47-4ae7-ffaa-001fa9b932fe"
      },
      "source": [
        "# representation of model layers\n",
        "from keras.utils import plot_model\n",
        "plot_model(mlp, to_file='plots/mlp_binary.png', show_shapes=True,)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 151
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZOUNj1J8Hhzo"
      },
      "source": [
        "# **Long Short-Term Memory Classifier (Binary Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7Twh09eaIMIs"
      },
      "source": [
        "# importing library for LSTM layers\n",
        "from keras.layers import LSTM"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0NuShoBumsdn"
      },
      "source": [
        "X = bin_data.iloc[:,0:93] # dataset excluding target attribute (encoded, one-hot-encoded,original)\n",
        "Y = bin_data[['intrusion']].values # target attribute"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rlH9mkLimrYW"
      },
      "source": [
        "# splitting the dataset 75% for training and 25% testing\n",
        "X_train, X_test, y_train, y_test = train_test_split(X,Y, test_size=0.25, random_state=42)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "15q5ptxAIWGt"
      },
      "source": [
        "X_train = X_train.to_numpy()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nJA6scawIW6U"
      },
      "source": [
        "x_train = np.reshape(X_train, (X_train.shape[0],X_train.shape[1],1))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "o8piatdjHo4w"
      },
      "source": [
        "lst = Sequential() # initializing model\n",
        "\n",
        "# input layer and LSTM layer with 50 neurons\n",
        "lst.add(LSTM(units=50, return_sequences=True, input_shape=(x_train.shape[1],1)))\n",
        "# outpute layer with sigmoid activation\n",
        "lst.add(Dense(1, activation='sigmoid'))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6ekKF6NFH_Iw"
      },
      "source": [
        "# defining loss function, optimizer, metrics and then compiling model\n",
        "lst.compile(loss='binary_crossentropy',optimizer='adam',metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nXF6pt7JlKxO",
        "outputId": "ec196c0a-72bf-4cc0-b8a3-95feff47cdb9"
      },
      "source": [
        "# summary of model layers\n",
        "lst.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "lstm_1 (LSTM)                (None, 93, 50)            10400     \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 93, 1)             51        \n",
            "=================================================================\n",
            "Total params: 10,451\n",
            "Trainable params: 10,451\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "peataArRecpK",
        "outputId": "01a36ce0-163b-4ac3-c0bf-fd6c0c4a608b"
      },
      "source": [
        "# training the model on training dataset\n",
        "history = lst.fit(x_train, y_train, epochs=100, batch_size=5000,validation_split=0.2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "16/16 [==============================] - 29s 2s/step - loss: 0.6915 - accuracy: 0.5119 - val_loss: 0.6863 - val_accuracy: 0.5798\n",
            "Epoch 2/100\n",
            "16/16 [==============================] - 26s 2s/step - loss: 0.6487 - accuracy: 0.6194 - val_loss: 0.4693 - val_accuracy: 0.8163\n",
            "Epoch 3/100\n",
            "16/16 [==============================] - 25s 2s/step - loss: 0.4611 - accuracy: 0.8119 - val_loss: 0.4278 - val_accuracy: 0.8231\n",
            "Epoch 4/100\n",
            "16/16 [==============================] - 25s 2s/step - loss: 0.4374 - accuracy: 0.8158 - val_loss: 0.4351 - val_accuracy: 0.8118\n",
            "Epoch 5/100\n",
            "16/16 [==============================] - 25s 2s/step - loss: 0.4359 - accuracy: 0.8105 - val_loss: 0.4117 - val_accuracy: 0.8281\n",
            "Epoch 6/100\n",
            "16/16 [==============================] - 25s 2s/step - loss: 0.4162 - accuracy: 0.8260 - val_loss: 0.4076 - val_accuracy: 0.8300\n",
            "Epoch 7/100\n",
            "16/16 [==============================] - 25s 2s/step - loss: 0.4156 - accuracy: 0.8265 - val_loss: 0.4064 - val_accuracy: 0.8305\n",
            "Epoch 8/100\n",
            "16/16 [==============================] - 26s 2s/step - loss: 0.4146 - accuracy: 0.8263 - val_loss: 0.4050 - val_accuracy: 0.8308\n",
            "Epoch 9/100\n",
            "16/16 [==============================] - 26s 2s/step - loss: 0.4108 - accuracy: 0.8274 - val_loss: 0.3915 - val_accuracy: 0.8318\n",
            "Epoch 10/100\n",
            "16/16 [==============================] - 25s 2s/step - loss: 0.3847 - accuracy: 0.8281 - val_loss: 0.3541 - val_accuracy: 0.8320\n",
            "Epoch 11/100\n",
            "16/16 [==============================] - 25s 2s/step - loss: 0.3520 - accuracy: 0.8276 - val_loss: 0.3428 - val_accuracy: 0.8321\n",
            "Epoch 12/100\n",
            "16/16 [==============================] - 25s 2s/step - loss: 0.3366 - accuracy: 0.8290 - val_loss: 0.3394 - val_accuracy: 0.7968\n",
            "Epoch 13/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3318 - accuracy: 0.8095 - val_loss: 0.3101 - val_accuracy: 0.8496\n",
            "Epoch 14/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3196 - accuracy: 0.8279 - val_loss: 0.3273 - val_accuracy: 0.7995\n",
            "Epoch 15/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3397 - accuracy: 0.8185 - val_loss: 0.4559 - val_accuracy: 0.7544\n",
            "Epoch 16/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.4291 - accuracy: 0.7626 - val_loss: 0.3367 - val_accuracy: 0.8271\n",
            "Epoch 17/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3351 - accuracy: 0.8235 - val_loss: 0.3342 - val_accuracy: 0.8301\n",
            "Epoch 18/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3505 - accuracy: 0.8320 - val_loss: 0.3216 - val_accuracy: 0.8756\n",
            "Epoch 19/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3276 - accuracy: 0.8794 - val_loss: 0.3200 - val_accuracy: 0.8839\n",
            "Epoch 20/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3256 - accuracy: 0.8812 - val_loss: 0.3184 - val_accuracy: 0.8793\n",
            "Epoch 21/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3191 - accuracy: 0.8837 - val_loss: 0.3162 - val_accuracy: 0.8825\n",
            "Epoch 22/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3165 - accuracy: 0.8858 - val_loss: 0.3149 - val_accuracy: 0.8816\n",
            "Epoch 23/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3180 - accuracy: 0.8831 - val_loss: 0.3072 - val_accuracy: 0.8826\n",
            "Epoch 24/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3003 - accuracy: 0.8852 - val_loss: 0.2902 - val_accuracy: 0.8832\n",
            "Epoch 25/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.2899 - accuracy: 0.8854 - val_loss: 0.2931 - val_accuracy: 0.8835\n",
            "Epoch 26/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.2885 - accuracy: 0.8863 - val_loss: 0.2903 - val_accuracy: 0.8821\n",
            "Epoch 27/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.2930 - accuracy: 0.8796 - val_loss: 0.3103 - val_accuracy: 0.8819\n",
            "Epoch 28/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3068 - accuracy: 0.8831 - val_loss: 0.3127 - val_accuracy: 0.8631\n",
            "Epoch 29/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3103 - accuracy: 0.8656 - val_loss: 0.3079 - val_accuracy: 0.8703\n",
            "Epoch 30/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3017 - accuracy: 0.8773 - val_loss: 0.2905 - val_accuracy: 0.8810\n",
            "Epoch 31/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.2882 - accuracy: 0.8847 - val_loss: 0.2871 - val_accuracy: 0.8835\n",
            "Epoch 32/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.2965 - accuracy: 0.8828 - val_loss: 0.3060 - val_accuracy: 0.8759\n",
            "Epoch 33/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.2952 - accuracy: 0.8820 - val_loss: 0.3054 - val_accuracy: 0.8842\n",
            "Epoch 34/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3048 - accuracy: 0.8842 - val_loss: 0.2881 - val_accuracy: 0.8813\n",
            "Epoch 35/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.2847 - accuracy: 0.8854 - val_loss: 0.3601 - val_accuracy: 0.8413\n",
            "Epoch 36/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.4467 - accuracy: 0.8329 - val_loss: 0.4251 - val_accuracy: 0.8320\n",
            "Epoch 37/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.4209 - accuracy: 0.8290 - val_loss: 0.3997 - val_accuracy: 0.8348\n",
            "Epoch 38/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.4066 - accuracy: 0.8313 - val_loss: 0.3984 - val_accuracy: 0.8378\n",
            "Epoch 39/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.4023 - accuracy: 0.8357 - val_loss: 0.3921 - val_accuracy: 0.8411\n",
            "Epoch 40/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3916 - accuracy: 0.8434 - val_loss: 0.3378 - val_accuracy: 0.8826\n",
            "Epoch 41/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3374 - accuracy: 0.8813 - val_loss: 0.3244 - val_accuracy: 0.8781\n",
            "Epoch 42/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3235 - accuracy: 0.8807 - val_loss: 0.3173 - val_accuracy: 0.8797\n",
            "Epoch 43/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3186 - accuracy: 0.8822 - val_loss: 0.3149 - val_accuracy: 0.8834\n",
            "Epoch 44/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3172 - accuracy: 0.8845 - val_loss: 0.3175 - val_accuracy: 0.8785\n",
            "Epoch 45/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3201 - accuracy: 0.8790 - val_loss: 0.3171 - val_accuracy: 0.8769\n",
            "Epoch 46/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3177 - accuracy: 0.8807 - val_loss: 0.3162 - val_accuracy: 0.8789\n",
            "Epoch 47/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3188 - accuracy: 0.8804 - val_loss: 0.3153 - val_accuracy: 0.8795\n",
            "Epoch 48/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3160 - accuracy: 0.8825 - val_loss: 0.3120 - val_accuracy: 0.8832\n",
            "Epoch 49/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3154 - accuracy: 0.8832 - val_loss: 0.3090 - val_accuracy: 0.8838\n",
            "Epoch 50/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3122 - accuracy: 0.8841 - val_loss: 0.2955 - val_accuracy: 0.8845\n",
            "Epoch 51/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.2942 - accuracy: 0.8856 - val_loss: 0.2884 - val_accuracy: 0.8855\n",
            "Epoch 52/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.2921 - accuracy: 0.8851 - val_loss: 0.2868 - val_accuracy: 0.8851\n",
            "Epoch 53/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.2879 - accuracy: 0.8854 - val_loss: 0.2825 - val_accuracy: 0.8865\n",
            "Epoch 54/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.2863 - accuracy: 0.8868 - val_loss: 0.2915 - val_accuracy: 0.8806\n",
            "Epoch 55/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.2924 - accuracy: 0.8824 - val_loss: 0.2905 - val_accuracy: 0.8864\n",
            "Epoch 56/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.2942 - accuracy: 0.8854 - val_loss: 0.3004 - val_accuracy: 0.8831\n",
            "Epoch 57/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.2946 - accuracy: 0.8868 - val_loss: 0.2938 - val_accuracy: 0.8857\n",
            "Epoch 58/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3071 - accuracy: 0.8669 - val_loss: 0.3347 - val_accuracy: 0.8330\n",
            "Epoch 59/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3307 - accuracy: 0.8297 - val_loss: 0.3176 - val_accuracy: 0.8336\n",
            "Epoch 60/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3132 - accuracy: 0.8397 - val_loss: 0.3026 - val_accuracy: 0.8546\n",
            "Epoch 61/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3144 - accuracy: 0.8382 - val_loss: 0.3056 - val_accuracy: 0.8400\n",
            "Epoch 62/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3094 - accuracy: 0.8358 - val_loss: 0.3042 - val_accuracy: 0.8646\n",
            "Epoch 63/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3036 - accuracy: 0.8599 - val_loss: 0.3266 - val_accuracy: 0.8117\n",
            "Epoch 64/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3421 - accuracy: 0.8478 - val_loss: 0.4597 - val_accuracy: 0.8331\n",
            "Epoch 65/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.4486 - accuracy: 0.8279 - val_loss: 0.3927 - val_accuracy: 0.8376\n",
            "Epoch 66/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.4004 - accuracy: 0.8341 - val_loss: 0.3942 - val_accuracy: 0.8391\n",
            "Epoch 67/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3977 - accuracy: 0.8362 - val_loss: 0.3895 - val_accuracy: 0.8391\n",
            "Epoch 68/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3919 - accuracy: 0.8387 - val_loss: 0.3876 - val_accuracy: 0.8403\n",
            "Epoch 69/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3932 - accuracy: 0.8375 - val_loss: 0.3836 - val_accuracy: 0.8432\n",
            "Epoch 70/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3817 - accuracy: 0.8459 - val_loss: 0.3314 - val_accuracy: 0.8852\n",
            "Epoch 71/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3256 - accuracy: 0.8852 - val_loss: 0.3098 - val_accuracy: 0.8832\n",
            "Epoch 72/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3118 - accuracy: 0.8856 - val_loss: 0.3081 - val_accuracy: 0.8847\n",
            "Epoch 73/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3109 - accuracy: 0.8868 - val_loss: 0.3085 - val_accuracy: 0.8930\n",
            "Epoch 74/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3107 - accuracy: 0.8922 - val_loss: 0.3065 - val_accuracy: 0.8915\n",
            "Epoch 75/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3068 - accuracy: 0.8934 - val_loss: 0.3087 - val_accuracy: 0.8898\n",
            "Epoch 76/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3120 - accuracy: 0.8897 - val_loss: 0.3037 - val_accuracy: 0.8941\n",
            "Epoch 77/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3087 - accuracy: 0.8920 - val_loss: 0.3064 - val_accuracy: 0.8899\n",
            "Epoch 78/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3068 - accuracy: 0.8915 - val_loss: 0.3040 - val_accuracy: 0.8902\n",
            "Epoch 79/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3031 - accuracy: 0.8927 - val_loss: 0.3025 - val_accuracy: 0.8902\n",
            "Epoch 80/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.2999 - accuracy: 0.8937 - val_loss: 0.3034 - val_accuracy: 0.8903\n",
            "Epoch 81/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3049 - accuracy: 0.8913 - val_loss: 0.3029 - val_accuracy: 0.8915\n",
            "Epoch 82/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3036 - accuracy: 0.8938 - val_loss: 0.3010 - val_accuracy: 0.8931\n",
            "Epoch 83/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3037 - accuracy: 0.8939 - val_loss: 0.3009 - val_accuracy: 0.8943\n",
            "Epoch 84/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3016 - accuracy: 0.8954 - val_loss: 0.3063 - val_accuracy: 0.8894\n",
            "Epoch 85/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3048 - accuracy: 0.8921 - val_loss: 0.3038 - val_accuracy: 0.8913\n",
            "Epoch 86/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3052 - accuracy: 0.8931 - val_loss: 0.3022 - val_accuracy: 0.8924\n",
            "Epoch 87/100\n",
            "16/16 [==============================] - 27s 2s/step - loss: 0.3035 - accuracy: 0.8936 - val_loss: 0.3010 - val_accuracy: 0.8918\n",
            "Epoch 88/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3043 - accuracy: 0.8922 - val_loss: 0.3011 - val_accuracy: 0.8918\n",
            "Epoch 89/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3002 - accuracy: 0.8938 - val_loss: 0.3038 - val_accuracy: 0.8882\n",
            "Epoch 90/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3103 - accuracy: 0.8829 - val_loss: 0.3093 - val_accuracy: 0.8597\n",
            "Epoch 91/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.2955 - accuracy: 0.8646 - val_loss: 0.2801 - val_accuracy: 0.8839\n",
            "Epoch 92/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.2940 - accuracy: 0.8744 - val_loss: 0.2826 - val_accuracy: 0.8840\n",
            "Epoch 93/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.2805 - accuracy: 0.8809 - val_loss: 0.2775 - val_accuracy: 0.8902\n",
            "Epoch 94/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.2936 - accuracy: 0.8532 - val_loss: 0.2942 - val_accuracy: 0.8324\n",
            "Epoch 95/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.2961 - accuracy: 0.8536 - val_loss: 0.2924 - val_accuracy: 0.8628\n",
            "Epoch 96/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.2904 - accuracy: 0.8689 - val_loss: 0.2958 - val_accuracy: 0.8560\n",
            "Epoch 97/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.2973 - accuracy: 0.8569 - val_loss: 0.2927 - val_accuracy: 0.8688\n",
            "Epoch 98/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3022 - accuracy: 0.8584 - val_loss: 0.3277 - val_accuracy: 0.7922\n",
            "Epoch 99/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3227 - accuracy: 0.8049 - val_loss: 0.3113 - val_accuracy: 0.8344\n",
            "Epoch 100/100\n",
            "16/16 [==============================] - 28s 2s/step - loss: 0.3115 - accuracy: 0.8293 - val_loss: 0.3079 - val_accuracy: 0.8341\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "F7RKBuhZpM1s"
      },
      "source": [
        "x_test = np.reshape(X_test, (X_test.shape[0],X_test.shape[1],1))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "wEvPvD0N1BoA",
        "outputId": "6ee7d239-4426-466c-92b4-5855b5227099"
      },
      "source": [
        "filepath = './models/lst_binary.json'\n",
        "weightspath = './weights/lst_binary.h5'\n",
        "if (not path.isfile(filepath)):\n",
        "  # serialize model to JSON\n",
        "  lst_json = lst.to_json()\n",
        "  with open(filepath, \"w\") as json_file:\n",
        "    json_file.write(lst_json)\n",
        "\n",
        "  # serialize weights to HDF5\n",
        "  lst.save_weights(weightspath)\n",
        "  print(\"Saved model to disk\")\n",
        "\n",
        "  # load json and create model\n",
        "  json_file = open(filepath, 'r')\n",
        "  loaded_model_json = json_file.read()\n",
        "  json_file.close()\n",
        "  lst = model_from_json(loaded_model_json)\n",
        "\n",
        "  # load weights into new model\n",
        "  lst.load_weights(weightspath)\n",
        "  print(\"Loaded model from disk\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ElLuZrYc1PCC"
      },
      "source": [
        "# defining loss function, optimizer, metrics and then compiling model\n",
        "lst.compile(loss='binary_crossentropy',optimizer='adam',metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "h0x7huDdpA0q",
        "outputId": "8c0a8451-9056-454b-aec4-2f51f0adfcb5"
      },
      "source": [
        "# predicting target attribute on testing dataset\n",
        "test_results = lst.evaluate(x_test, y_test, verbose=1)\n",
        "print(f'Test results - Loss: {test_results[0]} - Accuracy: {test_results[1]*100}%')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "985/985 [==============================] - 10s 10ms/step - loss: 0.3087 - accuracy: 0.8306\n",
            "Test results - Loss: 0.3086586892604828 - Accuracy: 83.05730819702148%\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "pJfKAHM0Ndpv",
        "outputId": "da1a6431-632a-4f6a-ce17-f93123f7dad9"
      },
      "source": [
        "# Plot of accuracy vs epoch of train and test dataset\n",
        "plt.plot(history.history['accuracy'])\n",
        "plt.plot(history.history['val_accuracy'])\n",
        "plt.title(\"Plot of accuracy vs epoch for train and test dataset\")\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='best')\n",
        "plt.savefig('plots/lstm_binary_accuracy.png')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "kd0qPmceNdAr",
        "outputId": "98ecf942-e282-4d63-9d15-af43d4bff193"
      },
      "source": [
        "# Plot of loss vs epoch of train and test dataset\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title(\"Plot of loss vs epoch for train and test dataset\")\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='best')\n",
        "plt.savefig('plots/lstm_binary_loss.png')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 312
        },
        "id": "RULmRxZ5zkkF",
        "outputId": "f8da3e6d-77d1-46ff-8808-aebc0a57c853"
      },
      "source": [
        "# representation of model layers\n",
        "plot_model(lst, to_file='plots/lstm_binary.png', show_shapes=True,)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 152
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "A2mFGsLLkzvr"
      },
      "source": [
        "# **Auto Encoder Classifier (Binary Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Jteo7lYCC9eV"
      },
      "source": [
        "from keras.layers import Input\n",
        "from keras.models import Model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7uaj0DBK6HkN"
      },
      "source": [
        "# splitting the dataset 75% for training and 25% testing\n",
        "X_train, X_test = train_test_split(bin_data, test_size=0.25, random_state=42)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-12iPEfcGzHi"
      },
      "source": [
        "# dataset excluding target attribute (encoded, one-hot-encoded,original)\n",
        "X_train = X_train.drop(['intrusion','abnormal','normal','label'],axis=1) \n",
        "\n",
        "y_test = X_test['intrusion'] # target attribute\n",
        "\n",
        "# dataset excluding target attribute (encoded, one-hot-encoded,original)\n",
        "X_test = X_test.drop(['intrusion','abnormal','normal','label'],axis=1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "347gsEWqAQHJ"
      },
      "source": [
        "X_train = X_train.values\n",
        "X_test = X_test.values\n",
        "y_test = y_test.values"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wxoezCMzlC0-"
      },
      "source": [
        "input_dim = X_train.shape[1]\n",
        "encoding_dim = 50\n",
        "\n",
        "#input layer\n",
        "input_layer = Input(shape=(input_dim, ))\n",
        "#encoding layer with 50 neurons\n",
        "encoder = Dense(encoding_dim, activation=\"relu\")(input_layer)           \n",
        "#decoding and output layer\n",
        "output_layer = Dense(input_dim, activation='softmax')(encoder)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SMS9roLx09rn"
      },
      "source": [
        "# creating model with input, encoding, decoding, output layers\n",
        "autoencoder = Model(inputs=input_layer, outputs=output_layer)\n",
        "\n",
        "# defining loss function, optimizer, metrics and then compiling model\n",
        "autoencoder.compile(optimizer='adam', loss='mean_squared_error',metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "oEBUVTQG9rMa",
        "outputId": "bd676303-d1dc-49b7-c246-2566950cbb51"
      },
      "source": [
        "# summary of model layers\n",
        "autoencoder.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_1 (InputLayer)         [(None, 93)]              0         \n",
            "_________________________________________________________________\n",
            "dense_4 (Dense)              (None, 50)                4700      \n",
            "_________________________________________________________________\n",
            "dense_5 (Dense)              (None, 93)                4743      \n",
            "=================================================================\n",
            "Total params: 9,443\n",
            "Trainable params: 9,443\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UChFX8029tsb",
        "outputId": "88a670c2-2872-4c35-a2fb-18ba5185336a"
      },
      "source": [
        "# training the model on training dataset\n",
        "history = autoencoder.fit(X_train, X_train, epochs=100,batch_size=500,validation_data=(X_test, X_test)).history"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1214 - accuracy: 0.0525 - val_loss: 0.1063 - val_accuracy: 0.4071\n",
            "Epoch 2/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1056 - accuracy: 0.6072 - val_loss: 0.1051 - val_accuracy: 0.6891\n",
            "Epoch 3/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1048 - accuracy: 0.7082 - val_loss: 0.1049 - val_accuracy: 0.7015\n",
            "Epoch 4/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1048 - accuracy: 0.7324 - val_loss: 0.1048 - val_accuracy: 0.7213\n",
            "Epoch 5/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.7678 - val_loss: 0.1047 - val_accuracy: 0.8029\n",
            "Epoch 6/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8180 - val_loss: 0.1047 - val_accuracy: 0.8578\n",
            "Epoch 7/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8400 - val_loss: 0.1047 - val_accuracy: 0.8321\n",
            "Epoch 8/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.8502 - val_loss: 0.1047 - val_accuracy: 0.8539\n",
            "Epoch 9/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8635 - val_loss: 0.1047 - val_accuracy: 0.8890\n",
            "Epoch 10/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8722 - val_loss: 0.1046 - val_accuracy: 0.8444\n",
            "Epoch 11/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.8681 - val_loss: 0.1046 - val_accuracy: 0.8839\n",
            "Epoch 12/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1045 - accuracy: 0.8747 - val_loss: 0.1046 - val_accuracy: 0.8875\n",
            "Epoch 13/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1044 - accuracy: 0.8776 - val_loss: 0.1046 - val_accuracy: 0.8489\n",
            "Epoch 14/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1042 - accuracy: 0.8744 - val_loss: 0.1046 - val_accuracy: 0.8619\n",
            "Epoch 15/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1044 - accuracy: 0.8765 - val_loss: 0.1046 - val_accuracy: 0.8391\n",
            "Epoch 16/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.8715 - val_loss: 0.1046 - val_accuracy: 0.8563\n",
            "Epoch 17/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8716 - val_loss: 0.1046 - val_accuracy: 0.8907\n",
            "Epoch 18/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1043 - accuracy: 0.8744 - val_loss: 0.1046 - val_accuracy: 0.9041\n",
            "Epoch 19/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8761 - val_loss: 0.1046 - val_accuracy: 0.8860\n",
            "Epoch 20/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1044 - accuracy: 0.8711 - val_loss: 0.1046 - val_accuracy: 0.8490\n",
            "Epoch 21/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1042 - accuracy: 0.8771 - val_loss: 0.1046 - val_accuracy: 0.9056\n",
            "Epoch 22/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.8786 - val_loss: 0.1046 - val_accuracy: 0.8829\n",
            "Epoch 23/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1043 - accuracy: 0.8769 - val_loss: 0.1046 - val_accuracy: 0.8751\n",
            "Epoch 24/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1045 - accuracy: 0.8795 - val_loss: 0.1046 - val_accuracy: 0.8784\n",
            "Epoch 25/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.8809 - val_loss: 0.1046 - val_accuracy: 0.8768\n",
            "Epoch 26/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8824 - val_loss: 0.1046 - val_accuracy: 0.8703\n",
            "Epoch 27/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8782 - val_loss: 0.1046 - val_accuracy: 0.8776\n",
            "Epoch 28/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.8723 - val_loss: 0.1046 - val_accuracy: 0.9076\n",
            "Epoch 29/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.8749 - val_loss: 0.1046 - val_accuracy: 0.8692\n",
            "Epoch 30/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8825 - val_loss: 0.1046 - val_accuracy: 0.8964\n",
            "Epoch 31/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.8895 - val_loss: 0.1046 - val_accuracy: 0.8889\n",
            "Epoch 32/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8868 - val_loss: 0.1046 - val_accuracy: 0.8901\n",
            "Epoch 33/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8891 - val_loss: 0.1046 - val_accuracy: 0.8924\n",
            "Epoch 34/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1044 - accuracy: 0.8890 - val_loss: 0.1046 - val_accuracy: 0.8868\n",
            "Epoch 35/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8915 - val_loss: 0.1046 - val_accuracy: 0.8787\n",
            "Epoch 36/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.8926 - val_loss: 0.1046 - val_accuracy: 0.8946\n",
            "Epoch 37/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8950 - val_loss: 0.1046 - val_accuracy: 0.9071\n",
            "Epoch 38/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.8936 - val_loss: 0.1046 - val_accuracy: 0.8880\n",
            "Epoch 39/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.8981 - val_loss: 0.1046 - val_accuracy: 0.8783\n",
            "Epoch 40/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.8967 - val_loss: 0.1046 - val_accuracy: 0.9142\n",
            "Epoch 41/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.8995 - val_loss: 0.1046 - val_accuracy: 0.9112\n",
            "Epoch 42/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.8972 - val_loss: 0.1046 - val_accuracy: 0.8945\n",
            "Epoch 43/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.8989 - val_loss: 0.1046 - val_accuracy: 0.9053\n",
            "Epoch 44/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.9008 - val_loss: 0.1046 - val_accuracy: 0.9423\n",
            "Epoch 45/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9011 - val_loss: 0.1046 - val_accuracy: 0.9121\n",
            "Epoch 46/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9046 - val_loss: 0.1046 - val_accuracy: 0.9086\n",
            "Epoch 47/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9010 - val_loss: 0.1046 - val_accuracy: 0.9046\n",
            "Epoch 48/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9018 - val_loss: 0.1046 - val_accuracy: 0.9101\n",
            "Epoch 49/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.9052 - val_loss: 0.1046 - val_accuracy: 0.9005\n",
            "Epoch 50/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.9021 - val_loss: 0.1046 - val_accuracy: 0.9002\n",
            "Epoch 51/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.9043 - val_loss: 0.1046 - val_accuracy: 0.9118\n",
            "Epoch 52/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.9061 - val_loss: 0.1046 - val_accuracy: 0.8976\n",
            "Epoch 53/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.9042 - val_loss: 0.1046 - val_accuracy: 0.9160\n",
            "Epoch 54/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1045 - accuracy: 0.9054 - val_loss: 0.1046 - val_accuracy: 0.9379\n",
            "Epoch 55/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1042 - accuracy: 0.9078 - val_loss: 0.1046 - val_accuracy: 0.9240\n",
            "Epoch 56/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9082 - val_loss: 0.1046 - val_accuracy: 0.8829\n",
            "Epoch 57/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.9063 - val_loss: 0.1046 - val_accuracy: 0.9355\n",
            "Epoch 58/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9100 - val_loss: 0.1046 - val_accuracy: 0.8864\n",
            "Epoch 59/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1044 - accuracy: 0.9086 - val_loss: 0.1046 - val_accuracy: 0.8953\n",
            "Epoch 60/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9072 - val_loss: 0.1046 - val_accuracy: 0.8933\n",
            "Epoch 61/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9077 - val_loss: 0.1046 - val_accuracy: 0.9145\n",
            "Epoch 62/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.9075 - val_loss: 0.1046 - val_accuracy: 0.9012\n",
            "Epoch 63/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9079 - val_loss: 0.1046 - val_accuracy: 0.9194\n",
            "Epoch 64/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9088 - val_loss: 0.1046 - val_accuracy: 0.9280\n",
            "Epoch 65/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1043 - accuracy: 0.9077 - val_loss: 0.1046 - val_accuracy: 0.9067\n",
            "Epoch 66/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9087 - val_loss: 0.1046 - val_accuracy: 0.9489\n",
            "Epoch 67/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.9096 - val_loss: 0.1046 - val_accuracy: 0.8849\n",
            "Epoch 68/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9094 - val_loss: 0.1046 - val_accuracy: 0.9129\n",
            "Epoch 69/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9091 - val_loss: 0.1046 - val_accuracy: 0.9039\n",
            "Epoch 70/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9099 - val_loss: 0.1046 - val_accuracy: 0.8840\n",
            "Epoch 71/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9056 - val_loss: 0.1046 - val_accuracy: 0.9125\n",
            "Epoch 72/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9093 - val_loss: 0.1046 - val_accuracy: 0.9081\n",
            "Epoch 73/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.9083 - val_loss: 0.1046 - val_accuracy: 0.9223\n",
            "Epoch 74/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9103 - val_loss: 0.1046 - val_accuracy: 0.9291\n",
            "Epoch 75/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9099 - val_loss: 0.1046 - val_accuracy: 0.9301\n",
            "Epoch 76/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9117 - val_loss: 0.1046 - val_accuracy: 0.9067\n",
            "Epoch 77/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.9092 - val_loss: 0.1046 - val_accuracy: 0.9443\n",
            "Epoch 78/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9094 - val_loss: 0.1046 - val_accuracy: 0.9180\n",
            "Epoch 79/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.9113 - val_loss: 0.1046 - val_accuracy: 0.9019\n",
            "Epoch 80/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.9115 - val_loss: 0.1046 - val_accuracy: 0.9522\n",
            "Epoch 81/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9105 - val_loss: 0.1046 - val_accuracy: 0.9054\n",
            "Epoch 82/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9097 - val_loss: 0.1046 - val_accuracy: 0.9070\n",
            "Epoch 83/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9082 - val_loss: 0.1046 - val_accuracy: 0.9220\n",
            "Epoch 84/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9112 - val_loss: 0.1046 - val_accuracy: 0.8926\n",
            "Epoch 85/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.9097 - val_loss: 0.1046 - val_accuracy: 0.8871\n",
            "Epoch 86/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9104 - val_loss: 0.1046 - val_accuracy: 0.9107\n",
            "Epoch 87/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9107 - val_loss: 0.1046 - val_accuracy: 0.8937\n",
            "Epoch 88/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9124 - val_loss: 0.1046 - val_accuracy: 0.9075\n",
            "Epoch 89/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9115 - val_loss: 0.1046 - val_accuracy: 0.8935\n",
            "Epoch 90/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9131 - val_loss: 0.1046 - val_accuracy: 0.9289\n",
            "Epoch 91/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9101 - val_loss: 0.1046 - val_accuracy: 0.8831\n",
            "Epoch 92/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1043 - accuracy: 0.9119 - val_loss: 0.1046 - val_accuracy: 0.8963\n",
            "Epoch 93/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.9127 - val_loss: 0.1046 - val_accuracy: 0.9140\n",
            "Epoch 94/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9132 - val_loss: 0.1046 - val_accuracy: 0.9410\n",
            "Epoch 95/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9140 - val_loss: 0.1046 - val_accuracy: 0.8723\n",
            "Epoch 96/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9132 - val_loss: 0.1046 - val_accuracy: 0.8909\n",
            "Epoch 97/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.9140 - val_loss: 0.1046 - val_accuracy: 0.9291\n",
            "Epoch 98/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9119 - val_loss: 0.1046 - val_accuracy: 0.9319\n",
            "Epoch 99/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9135 - val_loss: 0.1046 - val_accuracy: 0.9421\n",
            "Epoch 100/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9156 - val_loss: 0.1046 - val_accuracy: 0.9227\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JSZ9mOeM1mc4",
        "outputId": "5aa44835-8e18-4b69-d9be-434dff3b0bf6"
      },
      "source": [
        "filepath = './models/ae_binary.json'\n",
        "weightspath = './weights/ae_binary.h5'\n",
        "if (not path.isfile(filepath)): \n",
        "  # serialize model to JSON\n",
        "  autoencoder_json = autoencoder.to_json()\n",
        "  with open(filepath, \"w\") as json_file:\n",
        "    json_file.write(autoencoder_json)\n",
        "\n",
        "  # serialize weights to HDF5\n",
        "  autoencoder.save_weights(weightspath)\n",
        "  print(\"Saved model to disk\")\n",
        "\n",
        "  # load json and create model\n",
        "  json_file = open(filepath, 'r')\n",
        "  loaded_model_json = json_file.read()\n",
        "  json_file.close()\n",
        "  autoencoder = model_from_json(loaded_model_json)\n",
        "\n",
        "  # load weights into new model\n",
        "  autoencoder.load_weights(weightspath)\n",
        "  print(\"Loaded model from disk\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K8yypm1Z1nHr"
      },
      "source": [
        "# defining loss function, optimizer, metrics and then compiling model\n",
        "autoencoder.compile(optimizer='adam', loss='mean_squared_error',metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Tec3S1UI0Auz",
        "outputId": "77493bd3-114b-4993-d8e2-f1499003beb3"
      },
      "source": [
        "# predicting target attribute on testing dataset\n",
        "test_results = autoencoder.evaluate(X_test, X_test, verbose=1)\n",
        "print(f'Test results - Loss: {test_results[0]} - Accuracy: {test_results[1]}%')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "985/985 [==============================] - 1s 1ms/step - loss: 0.1044 - accuracy: 0.9203\n",
            "Test results - Loss: 0.10457229614257812 - Accuracy: 0.9226519465446472%\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "2G1lDF19gUCA",
        "outputId": "5c1ad1cc-b623-4e63-cb87-8a7e94c2019c"
      },
      "source": [
        "# Plot of loss vs epoch of train and test dataset\n",
        "plt.plot(history['loss'])\n",
        "plt.plot(history['val_loss'])\n",
        "plt.title(\"Plot of loss vs epoch for train and test dataset\")\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='upper right')\n",
        "plt.savefig('plots/ae_binary_loss.png')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "nF9C3reggW_z",
        "outputId": "1c77973c-e82a-4ce1-fa30-f98fe676e8d2"
      },
      "source": [
        "# Plot of accuracy vs epoch of train and test dataset\n",
        "plt.plot(history['accuracy'])\n",
        "plt.plot(history['val_accuracy'])\n",
        "plt.title(\"Plot of accuracy vs epoch for train and test dataset\")\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='best')\n",
        "plt.savefig('plots/ae_binary_accuracy.png')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 312
        },
        "id": "ZY2NaxgdEHSE",
        "outputId": "04e50955-eec4-44de-c1a5-decc620eca82"
      },
      "source": [
        "# representation of model layers\n",
        "plot_model(autoencoder, to_file='plots/ae_binary.png', show_shapes=True,)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 166
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lH62ub1mz98Z"
      },
      "source": [
        "# calculating reconstruction error \n",
        "predictions = autoencoder.predict(X_test)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_4-pGpAJ4UbZ"
      },
      "source": [
        "mse = np.mean(np.power(X_test - predictions, 2), axis=1)\n",
        "error_df = pd.DataFrame({'reconstruction_error': mse,\n",
        "                        'true_class': y_test})"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 297
        },
        "id": "V7eby65W4kHn",
        "outputId": "17456604-fe71-497c-f1cd-bc0225fdf422"
      },
      "source": [
        "error_df.describe()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>reconstruction_error</th>\n",
              "      <th>true_class</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>count</th>\n",
              "      <td>31494.000000</td>\n",
              "      <td>31494.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>mean</th>\n",
              "      <td>0.104572</td>\n",
              "      <td>0.532609</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>std</th>\n",
              "      <td>0.043577</td>\n",
              "      <td>0.498943</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>min</th>\n",
              "      <td>0.040633</td>\n",
              "      <td>0.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25%</th>\n",
              "      <td>0.074506</td>\n",
              "      <td>0.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50%</th>\n",
              "      <td>0.083821</td>\n",
              "      <td>1.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75%</th>\n",
              "      <td>0.151405</td>\n",
              "      <td>1.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>max</th>\n",
              "      <td>0.229802</td>\n",
              "      <td>1.000000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "       reconstruction_error    true_class\n",
              "count          31494.000000  31494.000000\n",
              "mean               0.104572      0.532609\n",
              "std                0.043577      0.498943\n",
              "min                0.040633      0.000000\n",
              "25%                0.074506      0.000000\n",
              "50%                0.083821      1.000000\n",
              "75%                0.151405      1.000000\n",
              "max                0.229802      1.000000"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 169
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GewtqonkCIgS"
      },
      "source": [
        "# **Linear Support Vector Machine Classifier (Multi-class Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6h3UrLZkyZ6F"
      },
      "source": [
        "X = multi_data.iloc[:,0:93].to_numpy() # dataset excluding target attribute (encoded, one-hot-encoded, original)\n",
        "Y = multi_data['intrusion'] # target attribute"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eciLJ53WycVR"
      },
      "source": [
        "# splitting the dataset 75% for training and 25% testing\n",
        "X_train, X_test, y_train, y_test = train_test_split(X,Y, test_size=0.25, random_state=42)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Puaavrt7Cpb1",
        "outputId": "68568619-cd09-41b0-fe72-21edc89c4728"
      },
      "source": [
        "lsvm=SVC(kernel='linear',gamma='auto')\n",
        "lsvm.fit(X_train,y_train) # training model on training dataset"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "SVC(C=1.0, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,\n",
              "    decision_function_shape='ovr', degree=3, gamma='auto', kernel='linear',\n",
              "    max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
              "    tol=0.001, verbose=False)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 172
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ju9JqNXAGBFa",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d733d435-bc9c-460d-9580-821b169b3c62"
      },
      "source": [
        "# saving trained model to disk\n",
        "pkl_filename = \"./models/lsvm_multi.pkl\"\n",
        "if (not path.isfile(pkl_filename)):\n",
        "  with open(pkl_filename, 'wb') as file:\n",
        "    pickle.dump(lsvm, file)\n",
        "  print(\"Saved model to disk\")\n",
        "  # loading trained model from disk\n",
        "  with open(pkl_filename, 'rb') as file:\n",
        "    lsvm = pickle.load(file)\n",
        "  print(\"Loade model from disk\")\n",
        "  "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loade model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yGGqsHE9ConV",
        "outputId": "e6c7c391-e2f9-4850-a0ac-53c5e3d7b4c8"
      },
      "source": [
        "y_pred=lsvm.predict(X_test) # predicting target attribute on testing dataset\n",
        "ac=accuracy_score(y_test, y_pred)*100  # calculating accuracy of predicted data\n",
        "print(\"LSVM-Classifier Multi-class Set-Accuracy is \", ac)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "LSVM-Classifier Multi-class Set-Accuracy is  95.24671365974471\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XG2oyW5PjPIP",
        "outputId": "333d9df0-8200-4f29-f9c5-ea6bde6bf74d"
      },
      "source": [
        "# classification report\n",
        "print(classification_report(y_test, y_pred,target_names=le2_classes_))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "         Dos       0.95      0.96      0.96     11484\n",
            "       Probe       0.86      0.79      0.82      2947\n",
            "         R2L       0.61      0.60      0.60       274\n",
            "         U2R       0.00      0.00      0.00        15\n",
            "      normal       0.97      0.98      0.98     16774\n",
            "\n",
            "    accuracy                           0.95     31494\n",
            "   macro avg       0.68      0.67      0.67     31494\n",
            "weighted avg       0.95      0.95      0.95     31494\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1272: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3M2ESEqWPqY4"
      },
      "source": [
        "# **Quadratic Support Vector Machine Classifier (Multi-class Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "h6iRNxewPkm1",
        "outputId": "205a15e0-7eb8-49f9-f3c3-3e97b17de6ae"
      },
      "source": [
        "qsvm=SVC(kernel='poly',gamma='auto')\n",
        "qsvm.fit(X_train,y_train) # training model on training dataset"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "SVC(C=1.0, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,\n",
              "    decision_function_shape='ovr', degree=3, gamma='auto', kernel='poly',\n",
              "    max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
              "    tol=0.001, verbose=False)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 176
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r-G-RZRtGUWs",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "22f748cc-cc4f-47f3-a2a2-05338dc20e55"
      },
      "source": [
        "pkl_filename = \"./models/qsvm_multi.pkl\"\n",
        "if (not path.isfile(pkl_filename)):\n",
        "  # saving trained model to disk\n",
        "  with open(pkl_filename, 'wb') as file:\n",
        "    pickle.dump(qsvm, file)\n",
        "  print(\"Saved model to disk\")\n",
        "  # loading trained model from disk\n",
        "  with open(pkl_filename, 'rb') as file:\n",
        "    qsvm = pickle.load(file)\n",
        "  print(\"Loaded model from disk\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YyiXVxb1PnER",
        "outputId": "441497be-d703-44ce-eea5-b3f77f8f1582"
      },
      "source": [
        "y_pred=qsvm.predict(X_test)  # predicting target attribute on testing dataset\n",
        "ac=accuracy_score(y_test, y_pred)*100  # calculating accuracy of predicted data\n",
        "print(\"QSVM-Classifier Multi-class Set-Accuracy is \", ac)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "QSVM-Classifier Multi-class Set-Accuracy is  92.86213246967677\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "mqVIJvYPjSUq",
        "outputId": "d46f5b41-934f-41c9-8804-2a1bd93cf46d"
      },
      "source": [
        "# classification report\n",
        "print(classification_report(y_test, y_pred,target_names=le2_classes_))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "         Dos       0.96      0.94      0.95     11484\n",
            "       Probe       0.96      0.61      0.74      2947\n",
            "         R2L       0.00      0.00      0.00       274\n",
            "         U2R       0.00      0.00      0.00        15\n",
            "      normal       0.91      1.00      0.95     16774\n",
            "\n",
            "    accuracy                           0.93     31494\n",
            "   macro avg       0.56      0.51      0.53     31494\n",
            "weighted avg       0.92      0.93      0.92     31494\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1272: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-HvlLolTuV8c"
      },
      "source": [
        "# **K-nearest-neighbor Classifier (Multi-class Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_oUD9fTyC5TQ",
        "outputId": "5f25a207-4d44-447d-8ee2-b36b70c46464"
      },
      "source": [
        "knn=KNeighborsClassifier(n_neighbors=5)\n",
        "knn.fit(X_train,y_train) # training model on training dataset"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',\n",
              "                     metric_params=None, n_jobs=None, n_neighbors=5, p=2,\n",
              "                     weights='uniform')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 180
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Jx8TF5VMGeM3",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d303a1d5-9788-40f4-93cc-9e3ceaf08f31"
      },
      "source": [
        "pkl_filename = \"./models/knn_multi.pkl\"\n",
        "if (not path.isfile(pkl_filename)):\n",
        "  # saving trained model to disk\n",
        "  with open(pkl_filename, 'wb') as file:\n",
        "    pickle.dump(knn, file)\n",
        "  print(\"Saved model to disk\")\n",
        "  # loading trained model from disk\n",
        "  with open(pkl_filename, 'rb') as file:\n",
        "    knn = pickle.load(file)\n",
        "  print(\"Loaded model from disk\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "S98xJiOLiqGp",
        "outputId": "d82ec7fa-270c-424d-865e-20b0ed20395d"
      },
      "source": [
        "y_pred=knn.predict(X_test)  # predicting target attribute on testing dataset\n",
        "ac=accuracy_score(y_test, y_pred)*100  # calculating accuracy of predicted data\n",
        "print(\"KNN-Classifier Multi-class Set-Accuracy is \", ac)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "KNN-Classifier Multi-class Set-Accuracy is  98.29173810884613\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bofv1igEjXEb",
        "outputId": "0a6d8de7-83dc-4d0b-f160-8ff908292b8b"
      },
      "source": [
        "# classification report\n",
        "print(classification_report(y_test, y_pred,target_names=le2_classes_))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "         Dos       0.99      0.99      0.99     11484\n",
            "       Probe       0.96      0.97      0.96      2947\n",
            "         R2L       0.92      0.87      0.89       274\n",
            "         U2R       0.40      0.13      0.20        15\n",
            "      normal       0.99      0.99      0.99     16774\n",
            "\n",
            "    accuracy                           0.98     31494\n",
            "   macro avg       0.85      0.79      0.81     31494\n",
            "weighted avg       0.98      0.98      0.98     31494\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cQWRqfCcJ1uu"
      },
      "source": [
        "# **Linear Discriminant Analysis Classifier (Multi-class Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Q-xx3vO4J086",
        "outputId": "f622fcbc-c4c8-4a97-ab24-95211c3b6879"
      },
      "source": [
        "lda = LinearDiscriminantAnalysis()\n",
        "lda.fit(X_train, y_train) # training model on training dataset"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "LinearDiscriminantAnalysis(n_components=None, priors=None, shrinkage=None,\n",
              "                           solver='svd', store_covariance=False, tol=0.0001)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 184
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6dkH8hMDGpjh",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "975706b2-e01e-4d1f-9186-22744f3f1969"
      },
      "source": [
        "pkl_filename = \"./models/lda_multi.pkl\"\n",
        "if (not path.isfile(pkl_filename)):\n",
        "  # saving trained model to disk\n",
        "  with open(pkl_filename, 'wb') as file:\n",
        "    pickle.dump(lda, file)\n",
        "  print(\"Saved model to disk\")\n",
        "  # loading trained model from disk\n",
        "  with open(pkl_filename, 'rb') as file:\n",
        "    lda = pickle.load(file)\n",
        "  print(\"Loaded model from disk\")\n",
        "  "
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "mk1MkfhjJ-AG",
        "outputId": "4b02935d-e8b2-40b5-c951-a4a213e5226e"
      },
      "source": [
        "y_pred = lda.predict(X_test)  # predicting target attribute on testing dataset\n",
        "ac=accuracy_score(y_test, y_pred)*100  # calculating accuracy of predicted data\n",
        "print(\"LDA-Classifier Set-Accuracy is \", ac)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "LDA-Classifier Set-Accuracy is  93.1923540991935\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "dhq5ifZsKLJW",
        "outputId": "e13bcb59-7f2a-45d0-9b75-59168c35b58e"
      },
      "source": [
        "# classification report\n",
        "print(classification_report(y_test, y_pred,target_names=le2_classes_))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "         Dos       0.94      0.96      0.95     11484\n",
            "       Probe       0.88      0.73      0.80      2947\n",
            "         R2L       0.37      0.89      0.52       274\n",
            "         U2R       0.03      0.47      0.06        15\n",
            "      normal       0.97      0.95      0.96     16774\n",
            "\n",
            "    accuracy                           0.93     31494\n",
            "   macro avg       0.64      0.80      0.66     31494\n",
            "weighted avg       0.94      0.93      0.94     31494\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Zn573OUHJONH"
      },
      "source": [
        "# **Quadratic Discriminant Analysis Classifier (Multi-class Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XXFXn3910kJA",
        "outputId": "419f861f-4d67-495f-c642-80beaaca75f3"
      },
      "source": [
        "qda = QuadraticDiscriminantAnalysis()\n",
        "qda.fit(X_train, y_train) # training model on training dataset"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/discriminant_analysis.py:691: UserWarning: Variables are collinear\n",
            "  warnings.warn(\"Variables are collinear\")\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "QuadraticDiscriminantAnalysis(priors=None, reg_param=0.0,\n",
              "                              store_covariance=False, tol=0.0001)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 188
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BeTtsaOzG1Ra",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "208c7771-25b5-46b8-9f35-9071ebd84a3f"
      },
      "source": [
        "pkl_filename = \"./models/qda_multi.pkl\"\n",
        "if (not path.isfile(pkl_filename)):\n",
        "  # saving trained model to disk\n",
        "  with open(pkl_filename, 'wb') as file:\n",
        "    pickle.dump(qda, file)\n",
        "  print(\"Saved model to disk\")\n",
        "  # loading trained model from disk\n",
        "  with open(pkl_filename, 'rb') as file:\n",
        "    qda = pickle.load(file)\n",
        "  print(\"Loaded model from disk\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zHCXEatcJZNj",
        "outputId": "ae5d7b60-03e8-494c-ca71-341e9eb24888"
      },
      "source": [
        "y_pred = qda.predict(X_test)  # predicting target attribute on testing dataset\n",
        "ac=accuracy_score(y_test, y_pred)*100  # calculating accuracy of predicted data\n",
        "print(\"QDA-Classifier Multi-class Set-Accuracy is \", ac)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "QDA-Classifier Multi-class Set-Accuracy is  44.96094494189369\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "oPEaIVtYJe-P",
        "outputId": "489bac6b-f8ab-4ba1-d5a7-ed5cb5ae8b3c"
      },
      "source": [
        "# classification report\n",
        "print(classification_report(y_test, y_pred,target_names=le2_classes_))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "         Dos       0.99      0.42      0.59     11484\n",
            "       Probe       0.97      0.06      0.11      2947\n",
            "         R2L       0.03      1.00      0.06       274\n",
            "         U2R       0.00      0.00      0.00        15\n",
            "      normal       0.50      0.53      0.51     16774\n",
            "\n",
            "    accuracy                           0.45     31494\n",
            "   macro avg       0.50      0.40      0.26     31494\n",
            "weighted avg       0.72      0.45      0.50     31494\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/metrics/_classification.py:1272: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "K5hU_51LKW-d"
      },
      "source": [
        "# **Multi Layer Perceptron Classifier (Multi-class Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9373ZC6VKX5G"
      },
      "source": [
        "X = multi_data.iloc[:,0:93]  # dataset excluding target attribute (encoded, one-hot-encoded,original)\n",
        "Y = multi_data[['Dos','normal','Probe','R2L','U2R']] # target attributes"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RfpZr6lyK3UG"
      },
      "source": [
        "# splitting the dataset 75% for training and 25% testing\n",
        "X_train, X_test, y_train, y_test = train_test_split(X,Y, test_size=0.25, random_state=42)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nlc90vl7LXqy"
      },
      "source": [
        "mlp = Sequential() # initializing model\n",
        "# input layer and first layer with 50 neurons\n",
        "mlp.add(Dense(units=50, input_dim=X_train.shape[1], activation='relu'))\n",
        "# output layer with softmax activation\n",
        "mlp.add(Dense(units=5,activation='softmax'))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sbjgI4YkLYFs"
      },
      "source": [
        "# defining loss function, optimizer, metrics and then compiling model\n",
        "mlp.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CAD9iBng6_bM",
        "outputId": "f2d021e7-6b23-4094-9f5a-215527ae4378"
      },
      "source": [
        "# summary of model layers\n",
        "mlp.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_3\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense_6 (Dense)              (None, 50)                4700      \n",
            "_________________________________________________________________\n",
            "dense_7 (Dense)              (None, 5)                 255       \n",
            "=================================================================\n",
            "Total params: 4,955\n",
            "Trainable params: 4,955\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0iR4f7BwLb9b",
        "outputId": "78fae018-4665-4187-d1d6-25f593660cb3"
      },
      "source": [
        "# training the model on training dataset\n",
        "history = mlp.fit(X_train, y_train, epochs=100, batch_size=5000,validation_split=0.2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "16/16 [==============================] - 1s 33ms/step - loss: 1.6282 - accuracy: 0.2066 - val_loss: 1.2330 - val_accuracy: 0.5749\n",
            "Epoch 2/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 1.1306 - accuracy: 0.6717 - val_loss: 0.8474 - val_accuracy: 0.8593\n",
            "Epoch 3/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.7775 - accuracy: 0.8580 - val_loss: 0.5909 - val_accuracy: 0.8665\n",
            "Epoch 4/100\n",
            "16/16 [==============================] - 0s 14ms/step - loss: 0.5489 - accuracy: 0.8655 - val_loss: 0.4435 - val_accuracy: 0.8720\n",
            "Epoch 5/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.4206 - accuracy: 0.8718 - val_loss: 0.3650 - val_accuracy: 0.8760\n",
            "Epoch 6/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.3533 - accuracy: 0.8797 - val_loss: 0.3201 - val_accuracy: 0.9093\n",
            "Epoch 7/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.3122 - accuracy: 0.9118 - val_loss: 0.2901 - val_accuracy: 0.9267\n",
            "Epoch 8/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.2817 - accuracy: 0.9273 - val_loss: 0.2679 - val_accuracy: 0.9336\n",
            "Epoch 9/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.2593 - accuracy: 0.9333 - val_loss: 0.2501 - val_accuracy: 0.9363\n",
            "Epoch 10/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.2423 - accuracy: 0.9365 - val_loss: 0.2353 - val_accuracy: 0.9390\n",
            "Epoch 11/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.2296 - accuracy: 0.9382 - val_loss: 0.2224 - val_accuracy: 0.9410\n",
            "Epoch 12/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.2155 - accuracy: 0.9433 - val_loss: 0.2113 - val_accuracy: 0.9457\n",
            "Epoch 13/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.2038 - accuracy: 0.9471 - val_loss: 0.2018 - val_accuracy: 0.9468\n",
            "Epoch 14/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1955 - accuracy: 0.9478 - val_loss: 0.1932 - val_accuracy: 0.9476\n",
            "Epoch 15/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.1891 - accuracy: 0.9489 - val_loss: 0.1857 - val_accuracy: 0.9499\n",
            "Epoch 16/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1827 - accuracy: 0.9499 - val_loss: 0.1791 - val_accuracy: 0.9501\n",
            "Epoch 17/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.1730 - accuracy: 0.9512 - val_loss: 0.1734 - val_accuracy: 0.9509\n",
            "Epoch 18/100\n",
            "16/16 [==============================] - 0s 14ms/step - loss: 0.1691 - accuracy: 0.9509 - val_loss: 0.1683 - val_accuracy: 0.9510\n",
            "Epoch 19/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1633 - accuracy: 0.9520 - val_loss: 0.1637 - val_accuracy: 0.9511\n",
            "Epoch 20/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.1602 - accuracy: 0.9515 - val_loss: 0.1595 - val_accuracy: 0.9515\n",
            "Epoch 21/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1553 - accuracy: 0.9521 - val_loss: 0.1558 - val_accuracy: 0.9516\n",
            "Epoch 22/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.1516 - accuracy: 0.9527 - val_loss: 0.1525 - val_accuracy: 0.9520\n",
            "Epoch 23/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1491 - accuracy: 0.9524 - val_loss: 0.1494 - val_accuracy: 0.9524\n",
            "Epoch 24/100\n",
            "16/16 [==============================] - 0s 15ms/step - loss: 0.1443 - accuracy: 0.9536 - val_loss: 0.1466 - val_accuracy: 0.9530\n",
            "Epoch 25/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1446 - accuracy: 0.9528 - val_loss: 0.1441 - val_accuracy: 0.9529\n",
            "Epoch 26/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1409 - accuracy: 0.9537 - val_loss: 0.1417 - val_accuracy: 0.9529\n",
            "Epoch 27/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1403 - accuracy: 0.9525 - val_loss: 0.1394 - val_accuracy: 0.9530\n",
            "Epoch 28/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1352 - accuracy: 0.9541 - val_loss: 0.1373 - val_accuracy: 0.9531\n",
            "Epoch 29/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1361 - accuracy: 0.9538 - val_loss: 0.1353 - val_accuracy: 0.9535\n",
            "Epoch 30/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1349 - accuracy: 0.9534 - val_loss: 0.1334 - val_accuracy: 0.9537\n",
            "Epoch 31/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1319 - accuracy: 0.9543 - val_loss: 0.1316 - val_accuracy: 0.9539\n",
            "Epoch 32/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.1305 - accuracy: 0.9541 - val_loss: 0.1299 - val_accuracy: 0.9544\n",
            "Epoch 33/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1276 - accuracy: 0.9548 - val_loss: 0.1282 - val_accuracy: 0.9549\n",
            "Epoch 34/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1264 - accuracy: 0.9562 - val_loss: 0.1267 - val_accuracy: 0.9561\n",
            "Epoch 35/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1257 - accuracy: 0.9564 - val_loss: 0.1252 - val_accuracy: 0.9569\n",
            "Epoch 36/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1219 - accuracy: 0.9579 - val_loss: 0.1238 - val_accuracy: 0.9575\n",
            "Epoch 37/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.1214 - accuracy: 0.9590 - val_loss: 0.1224 - val_accuracy: 0.9589\n",
            "Epoch 38/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1203 - accuracy: 0.9599 - val_loss: 0.1211 - val_accuracy: 0.9591\n",
            "Epoch 39/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1177 - accuracy: 0.9605 - val_loss: 0.1199 - val_accuracy: 0.9596\n",
            "Epoch 40/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.1162 - accuracy: 0.9607 - val_loss: 0.1188 - val_accuracy: 0.9602\n",
            "Epoch 41/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1170 - accuracy: 0.9614 - val_loss: 0.1177 - val_accuracy: 0.9609\n",
            "Epoch 42/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1160 - accuracy: 0.9601 - val_loss: 0.1166 - val_accuracy: 0.9613\n",
            "Epoch 43/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1148 - accuracy: 0.9616 - val_loss: 0.1156 - val_accuracy: 0.9614\n",
            "Epoch 44/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1146 - accuracy: 0.9618 - val_loss: 0.1146 - val_accuracy: 0.9616\n",
            "Epoch 45/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.1111 - accuracy: 0.9628 - val_loss: 0.1137 - val_accuracy: 0.9619\n",
            "Epoch 46/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1125 - accuracy: 0.9629 - val_loss: 0.1128 - val_accuracy: 0.9625\n",
            "Epoch 47/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1109 - accuracy: 0.9632 - val_loss: 0.1118 - val_accuracy: 0.9628\n",
            "Epoch 48/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1105 - accuracy: 0.9631 - val_loss: 0.1110 - val_accuracy: 0.9630\n",
            "Epoch 49/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1079 - accuracy: 0.9643 - val_loss: 0.1100 - val_accuracy: 0.9639\n",
            "Epoch 50/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1085 - accuracy: 0.9644 - val_loss: 0.1091 - val_accuracy: 0.9640\n",
            "Epoch 51/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.1087 - accuracy: 0.9640 - val_loss: 0.1083 - val_accuracy: 0.9642\n",
            "Epoch 52/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1061 - accuracy: 0.9647 - val_loss: 0.1077 - val_accuracy: 0.9641\n",
            "Epoch 53/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.1049 - accuracy: 0.9652 - val_loss: 0.1069 - val_accuracy: 0.9642\n",
            "Epoch 54/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1067 - accuracy: 0.9643 - val_loss: 0.1062 - val_accuracy: 0.9644\n",
            "Epoch 55/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1039 - accuracy: 0.9650 - val_loss: 0.1056 - val_accuracy: 0.9644\n",
            "Epoch 56/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1016 - accuracy: 0.9663 - val_loss: 0.1049 - val_accuracy: 0.9652\n",
            "Epoch 57/100\n",
            "16/16 [==============================] - 0s 15ms/step - loss: 0.1027 - accuracy: 0.9653 - val_loss: 0.1044 - val_accuracy: 0.9651\n",
            "Epoch 58/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1002 - accuracy: 0.9662 - val_loss: 0.1037 - val_accuracy: 0.9653\n",
            "Epoch 59/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1018 - accuracy: 0.9663 - val_loss: 0.1032 - val_accuracy: 0.9653\n",
            "Epoch 60/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.1006 - accuracy: 0.9665 - val_loss: 0.1026 - val_accuracy: 0.9656\n",
            "Epoch 61/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.1016 - accuracy: 0.9661 - val_loss: 0.1021 - val_accuracy: 0.9662\n",
            "Epoch 62/100\n",
            "16/16 [==============================] - 0s 14ms/step - loss: 0.0981 - accuracy: 0.9672 - val_loss: 0.1016 - val_accuracy: 0.9661\n",
            "Epoch 63/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.1013 - accuracy: 0.9663 - val_loss: 0.1011 - val_accuracy: 0.9664\n",
            "Epoch 64/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0984 - accuracy: 0.9675 - val_loss: 0.1007 - val_accuracy: 0.9664\n",
            "Epoch 65/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0972 - accuracy: 0.9676 - val_loss: 0.1003 - val_accuracy: 0.9664\n",
            "Epoch 66/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.0968 - accuracy: 0.9678 - val_loss: 0.0998 - val_accuracy: 0.9668\n",
            "Epoch 67/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0966 - accuracy: 0.9680 - val_loss: 0.0993 - val_accuracy: 0.9669\n",
            "Epoch 68/100\n",
            "16/16 [==============================] - 0s 14ms/step - loss: 0.0989 - accuracy: 0.9677 - val_loss: 0.0990 - val_accuracy: 0.9671\n",
            "Epoch 69/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0990 - accuracy: 0.9671 - val_loss: 0.0986 - val_accuracy: 0.9670\n",
            "Epoch 70/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.0967 - accuracy: 0.9680 - val_loss: 0.0981 - val_accuracy: 0.9672\n",
            "Epoch 71/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0975 - accuracy: 0.9684 - val_loss: 0.0979 - val_accuracy: 0.9675\n",
            "Epoch 72/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0963 - accuracy: 0.9678 - val_loss: 0.0973 - val_accuracy: 0.9677\n",
            "Epoch 73/100\n",
            "16/16 [==============================] - 0s 14ms/step - loss: 0.0934 - accuracy: 0.9691 - val_loss: 0.0970 - val_accuracy: 0.9678\n",
            "Epoch 74/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0938 - accuracy: 0.9690 - val_loss: 0.0966 - val_accuracy: 0.9681\n",
            "Epoch 75/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0940 - accuracy: 0.9690 - val_loss: 0.0963 - val_accuracy: 0.9682\n",
            "Epoch 76/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0937 - accuracy: 0.9689 - val_loss: 0.0959 - val_accuracy: 0.9681\n",
            "Epoch 77/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0919 - accuracy: 0.9693 - val_loss: 0.0957 - val_accuracy: 0.9682\n",
            "Epoch 78/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0943 - accuracy: 0.9688 - val_loss: 0.0953 - val_accuracy: 0.9682\n",
            "Epoch 79/100\n",
            "16/16 [==============================] - 0s 14ms/step - loss: 0.0930 - accuracy: 0.9694 - val_loss: 0.0952 - val_accuracy: 0.9683\n",
            "Epoch 80/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.0915 - accuracy: 0.9701 - val_loss: 0.0946 - val_accuracy: 0.9687\n",
            "Epoch 81/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0919 - accuracy: 0.9701 - val_loss: 0.0945 - val_accuracy: 0.9684\n",
            "Epoch 82/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0917 - accuracy: 0.9696 - val_loss: 0.0941 - val_accuracy: 0.9685\n",
            "Epoch 83/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0926 - accuracy: 0.9690 - val_loss: 0.0939 - val_accuracy: 0.9686\n",
            "Epoch 84/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.0913 - accuracy: 0.9700 - val_loss: 0.0935 - val_accuracy: 0.9689\n",
            "Epoch 85/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0896 - accuracy: 0.9700 - val_loss: 0.0934 - val_accuracy: 0.9688\n",
            "Epoch 86/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0910 - accuracy: 0.9698 - val_loss: 0.0930 - val_accuracy: 0.9689\n",
            "Epoch 87/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0907 - accuracy: 0.9699 - val_loss: 0.0929 - val_accuracy: 0.9686\n",
            "Epoch 88/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0903 - accuracy: 0.9702 - val_loss: 0.0924 - val_accuracy: 0.9693\n",
            "Epoch 89/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.0910 - accuracy: 0.9701 - val_loss: 0.0923 - val_accuracy: 0.9691\n",
            "Epoch 90/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0879 - accuracy: 0.9707 - val_loss: 0.0920 - val_accuracy: 0.9691\n",
            "Epoch 91/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0870 - accuracy: 0.9709 - val_loss: 0.0917 - val_accuracy: 0.9694\n",
            "Epoch 92/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0882 - accuracy: 0.9705 - val_loss: 0.0914 - val_accuracy: 0.9692\n",
            "Epoch 93/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.0883 - accuracy: 0.9704 - val_loss: 0.0913 - val_accuracy: 0.9691\n",
            "Epoch 94/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.0900 - accuracy: 0.9701 - val_loss: 0.0910 - val_accuracy: 0.9693\n",
            "Epoch 95/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0875 - accuracy: 0.9698 - val_loss: 0.0907 - val_accuracy: 0.9697\n",
            "Epoch 96/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0901 - accuracy: 0.9696 - val_loss: 0.0906 - val_accuracy: 0.9695\n",
            "Epoch 97/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0879 - accuracy: 0.9707 - val_loss: 0.0904 - val_accuracy: 0.9694\n",
            "Epoch 98/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0890 - accuracy: 0.9702 - val_loss: 0.0900 - val_accuracy: 0.9695\n",
            "Epoch 99/100\n",
            "16/16 [==============================] - 0s 12ms/step - loss: 0.0863 - accuracy: 0.9708 - val_loss: 0.0898 - val_accuracy: 0.9697\n",
            "Epoch 100/100\n",
            "16/16 [==============================] - 0s 13ms/step - loss: 0.0852 - accuracy: 0.9712 - val_loss: 0.0896 - val_accuracy: 0.9696\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sri14hur2zuN",
        "outputId": "9f948a59-930a-43da-c9ed-be7b67ce59bd"
      },
      "source": [
        "filepath = './models/mlp_multi.json'\n",
        "weightspath = './weights/mlp_multi.h5'\n",
        "if (not path.isfile(filepath)):\n",
        "  # serialize model to JSON\n",
        "  mlp_json = mlp.to_json()\n",
        "  with open(filepath, \"w\") as json_file:\n",
        "    json_file.write(mlp_json)\n",
        "\n",
        "  # serialize weights to HDF5\n",
        "  mlp.save_weights(weightspath)\n",
        "  print(\"Saved model to disk\")\n",
        "  \n",
        "  # load json and create model\n",
        "  json_file = open(filepath, 'r')\n",
        "  loaded_model_json = json_file.read()\n",
        "  json_file.close()\n",
        "  mlp = model_from_json(loaded_model_json)\n",
        "\n",
        "  # load weights into new model\n",
        "  mlp.load_weights(weightspath)\n",
        "  print(\"Loaded model from disk\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "H29GR38R21Bl"
      },
      "source": [
        "# defining loss function, optimizer, metrics and then compiling model\n",
        "mlp.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zkzQb_yHLhaX",
        "outputId": "27f6ad15-d9bc-45ce-8131-01e3d0e72063"
      },
      "source": [
        "# predicting target attribute on testing dataset\n",
        "test_results = mlp.evaluate(X_test, y_test, verbose=1)\n",
        "print(f'Test results - Loss: {test_results[0]} - Accuracy: {test_results[1]*100}%')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "985/985 [==============================] - 1s 936us/step - loss: 0.0941 - accuracy: 0.9682\n",
            "Test results - Loss: 0.0895298421382904 - Accuracy: 96.92639708518982%\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "znDf_uubLuH2",
        "outputId": "8c058731-edb5-41bb-bb5b-61f1905b3cda"
      },
      "source": [
        "# Plot of accuracy vs epoch for train and test dataset\n",
        "plt.plot(history.history['accuracy'])\n",
        "plt.plot(history.history['val_accuracy'])\n",
        "plt.title(\"Plot of accuracy vs epoch for train and test dataset\")\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='best')\n",
        "plt.savefig('plots/mlp_multi_accuracy.png')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "Ncfod0AWLw7W",
        "outputId": "6a907b11-cc54-4927-8d84-46c0fb05c73f"
      },
      "source": [
        "# Plot of loss vs epoch for train and test dataset\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title(\"Plot of loss vs epoch for train and test dataset\")\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='upper right')\n",
        "plt.savefig('plots/mlp_multi_loss.png')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 312
        },
        "id": "7AhHtm2r7T2-",
        "outputId": "cf54db0f-d769-400b-9b04-88f8c67f14f6"
      },
      "source": [
        "# representation of model layers\n",
        "plot_model(mlp, to_file='plots/mlp_multi.png', show_shapes=True,)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 203
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RpcZ4R0m8L7N"
      },
      "source": [
        "# **Auto Encoder Classifier (Multi-class Classification)**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VVNdyL8S8Vaw"
      },
      "source": [
        "# splitting the dataset 75% for training and 25% testing\n",
        "X_train, X_test = train_test_split(multi_data, test_size=0.25, random_state=42)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "L03tdsFq8Wdj"
      },
      "source": [
        "# dataset excluding target attribute (encoded, one-hot-encoded,original)\n",
        "X_train = X_train.drop(['intrusion','Dos','normal','Probe','R2L','U2R','label'],axis=1)\n",
        "\n",
        "y_test = X_test['intrusion'] # target attribute\n",
        "\n",
        "# dataset excluding target attribute (encoded, one-hot-encoded,original)\n",
        "X_test = X_test.drop(['intrusion','Dos','normal','Probe','R2L','U2R','label'],axis=1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4cT2cprf8bSf"
      },
      "source": [
        "X_train = X_train.values\n",
        "X_test = X_test.values\n",
        "y_test = y_test.values"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4eKXJ5JC8dWJ"
      },
      "source": [
        "input_dim = X_train.shape[1]\n",
        "encoding_dim = 50\n",
        "\n",
        "#input layer\n",
        "input_layer = Input(shape=(input_dim, ))\n",
        "#encoding layer with 50 neurons\n",
        "encoder = Dense(encoding_dim, activation=\"relu\")(input_layer)           \n",
        "#decoding and output layer\n",
        "output_layer = Dense(input_dim, activation='softmax')(encoder)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jWToOFFE4cEB"
      },
      "source": [
        "# creating model with input, encoding, decoding, output layers\n",
        "autoencoder = Model(inputs=input_layer, outputs=output_layer)\n",
        "\n",
        "# defining loss function, optimizer, metrics and then compiling model\n",
        "autoencoder.compile(optimizer='adam', loss='mean_squared_error',metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zKTV_xaK8ha8",
        "outputId": "52f998bd-01ec-4d25-e99d-9d49db04aea4"
      },
      "source": [
        "# summary of model layers\n",
        "autoencoder.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model_1\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_2 (InputLayer)         [(None, 93)]              0         \n",
            "_________________________________________________________________\n",
            "dense_8 (Dense)              (None, 50)                4700      \n",
            "_________________________________________________________________\n",
            "dense_9 (Dense)              (None, 93)                4743      \n",
            "=================================================================\n",
            "Total params: 9,443\n",
            "Trainable params: 9,443\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "dfO2sPjn8kCd",
        "outputId": "48d6a854-8a01-4daf-a375-955409e00edf"
      },
      "source": [
        "# training the model on training dataset\n",
        "history = autoencoder.fit(X_train, X_train, epochs=100,batch_size=500,validation_data=(X_test, X_test)).history"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1211 - accuracy: 0.1116 - val_loss: 0.1060 - val_accuracy: 0.3333\n",
            "Epoch 2/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1054 - accuracy: 0.4788 - val_loss: 0.1051 - val_accuracy: 0.6918\n",
            "Epoch 3/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1048 - accuracy: 0.6864 - val_loss: 0.1049 - val_accuracy: 0.6834\n",
            "Epoch 4/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.7274 - val_loss: 0.1048 - val_accuracy: 0.7741\n",
            "Epoch 5/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.7674 - val_loss: 0.1048 - val_accuracy: 0.8045\n",
            "Epoch 6/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.8088 - val_loss: 0.1047 - val_accuracy: 0.8379\n",
            "Epoch 7/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1047 - accuracy: 0.8389 - val_loss: 0.1047 - val_accuracy: 0.8338\n",
            "Epoch 8/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.8605 - val_loss: 0.1047 - val_accuracy: 0.8636\n",
            "Epoch 9/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.8686 - val_loss: 0.1047 - val_accuracy: 0.8626\n",
            "Epoch 10/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.8733 - val_loss: 0.1046 - val_accuracy: 0.8757\n",
            "Epoch 11/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.8749 - val_loss: 0.1046 - val_accuracy: 0.8832\n",
            "Epoch 12/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8783 - val_loss: 0.1046 - val_accuracy: 0.8705\n",
            "Epoch 13/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.8807 - val_loss: 0.1046 - val_accuracy: 0.8882\n",
            "Epoch 14/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8814 - val_loss: 0.1046 - val_accuracy: 0.8894\n",
            "Epoch 15/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8765 - val_loss: 0.1046 - val_accuracy: 0.8816\n",
            "Epoch 16/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1042 - accuracy: 0.8776 - val_loss: 0.1046 - val_accuracy: 0.8882\n",
            "Epoch 17/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8809 - val_loss: 0.1046 - val_accuracy: 0.8835\n",
            "Epoch 18/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1042 - accuracy: 0.8778 - val_loss: 0.1046 - val_accuracy: 0.9059\n",
            "Epoch 19/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1043 - accuracy: 0.8784 - val_loss: 0.1046 - val_accuracy: 0.8973\n",
            "Epoch 20/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8859 - val_loss: 0.1046 - val_accuracy: 0.8976\n",
            "Epoch 21/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1045 - accuracy: 0.8822 - val_loss: 0.1046 - val_accuracy: 0.8603\n",
            "Epoch 22/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1045 - accuracy: 0.8769 - val_loss: 0.1046 - val_accuracy: 0.8695\n",
            "Epoch 23/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1043 - accuracy: 0.8720 - val_loss: 0.1046 - val_accuracy: 0.8490\n",
            "Epoch 24/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8712 - val_loss: 0.1046 - val_accuracy: 0.8844\n",
            "Epoch 25/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1045 - accuracy: 0.8726 - val_loss: 0.1046 - val_accuracy: 0.8168\n",
            "Epoch 26/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1042 - accuracy: 0.8717 - val_loss: 0.1046 - val_accuracy: 0.9182\n",
            "Epoch 27/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8834 - val_loss: 0.1046 - val_accuracy: 0.8768\n",
            "Epoch 28/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1044 - accuracy: 0.8795 - val_loss: 0.1046 - val_accuracy: 0.8779\n",
            "Epoch 29/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.8802 - val_loss: 0.1046 - val_accuracy: 0.8919\n",
            "Epoch 30/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.8813 - val_loss: 0.1046 - val_accuracy: 0.8715\n",
            "Epoch 31/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8850 - val_loss: 0.1046 - val_accuracy: 0.8924\n",
            "Epoch 32/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8801 - val_loss: 0.1046 - val_accuracy: 0.8982\n",
            "Epoch 33/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8824 - val_loss: 0.1046 - val_accuracy: 0.8785\n",
            "Epoch 34/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.8803 - val_loss: 0.1046 - val_accuracy: 0.8947\n",
            "Epoch 35/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.8812 - val_loss: 0.1046 - val_accuracy: 0.9182\n",
            "Epoch 36/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8824 - val_loss: 0.1046 - val_accuracy: 0.9256\n",
            "Epoch 37/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1044 - accuracy: 0.8805 - val_loss: 0.1046 - val_accuracy: 0.9182\n",
            "Epoch 38/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.8789 - val_loss: 0.1046 - val_accuracy: 0.8771\n",
            "Epoch 39/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8778 - val_loss: 0.1046 - val_accuracy: 0.9007\n",
            "Epoch 40/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1041 - accuracy: 0.8858 - val_loss: 0.1046 - val_accuracy: 0.8668\n",
            "Epoch 41/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8817 - val_loss: 0.1046 - val_accuracy: 0.9312\n",
            "Epoch 42/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8862 - val_loss: 0.1046 - val_accuracy: 0.9068\n",
            "Epoch 43/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1043 - accuracy: 0.8879 - val_loss: 0.1046 - val_accuracy: 0.8892\n",
            "Epoch 44/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8835 - val_loss: 0.1046 - val_accuracy: 0.8528\n",
            "Epoch 45/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8863 - val_loss: 0.1046 - val_accuracy: 0.8876\n",
            "Epoch 46/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.8843 - val_loss: 0.1046 - val_accuracy: 0.8754\n",
            "Epoch 47/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1040 - accuracy: 0.8867 - val_loss: 0.1046 - val_accuracy: 0.8848\n",
            "Epoch 48/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8846 - val_loss: 0.1046 - val_accuracy: 0.8829\n",
            "Epoch 49/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1044 - accuracy: 0.8845 - val_loss: 0.1046 - val_accuracy: 0.8719\n",
            "Epoch 50/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8882 - val_loss: 0.1046 - val_accuracy: 0.9105\n",
            "Epoch 51/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.8899 - val_loss: 0.1046 - val_accuracy: 0.9427\n",
            "Epoch 52/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8933 - val_loss: 0.1046 - val_accuracy: 0.8755\n",
            "Epoch 53/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8904 - val_loss: 0.1046 - val_accuracy: 0.8981\n",
            "Epoch 54/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.8924 - val_loss: 0.1046 - val_accuracy: 0.8621\n",
            "Epoch 55/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1042 - accuracy: 0.8902 - val_loss: 0.1046 - val_accuracy: 0.9139\n",
            "Epoch 56/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.8947 - val_loss: 0.1046 - val_accuracy: 0.9121\n",
            "Epoch 57/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8964 - val_loss: 0.1046 - val_accuracy: 0.8844\n",
            "Epoch 58/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.8980 - val_loss: 0.1046 - val_accuracy: 0.9275\n",
            "Epoch 59/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1044 - accuracy: 0.8995 - val_loss: 0.1046 - val_accuracy: 0.9300\n",
            "Epoch 60/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9041 - val_loss: 0.1046 - val_accuracy: 0.8990\n",
            "Epoch 61/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.9032 - val_loss: 0.1046 - val_accuracy: 0.9065\n",
            "Epoch 62/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9052 - val_loss: 0.1046 - val_accuracy: 0.9297\n",
            "Epoch 63/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.9066 - val_loss: 0.1046 - val_accuracy: 0.8933\n",
            "Epoch 64/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9137 - val_loss: 0.1046 - val_accuracy: 0.9557\n",
            "Epoch 65/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9122 - val_loss: 0.1046 - val_accuracy: 0.8872\n",
            "Epoch 66/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.9109 - val_loss: 0.1046 - val_accuracy: 0.9039\n",
            "Epoch 67/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9150 - val_loss: 0.1046 - val_accuracy: 0.9266\n",
            "Epoch 68/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9166 - val_loss: 0.1046 - val_accuracy: 0.8617\n",
            "Epoch 69/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9117 - val_loss: 0.1046 - val_accuracy: 0.9324\n",
            "Epoch 70/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9167 - val_loss: 0.1046 - val_accuracy: 0.9322\n",
            "Epoch 71/100\n",
            "189/189 [==============================] - 1s 3ms/step - loss: 0.1043 - accuracy: 0.9182 - val_loss: 0.1046 - val_accuracy: 0.9087\n",
            "Epoch 72/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1039 - accuracy: 0.9185 - val_loss: 0.1046 - val_accuracy: 0.9133\n",
            "Epoch 73/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.9195 - val_loss: 0.1046 - val_accuracy: 0.9253\n",
            "Epoch 74/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9192 - val_loss: 0.1046 - val_accuracy: 0.9410\n",
            "Epoch 75/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9196 - val_loss: 0.1046 - val_accuracy: 0.9420\n",
            "Epoch 76/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9227 - val_loss: 0.1046 - val_accuracy: 0.8929\n",
            "Epoch 77/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.9206 - val_loss: 0.1046 - val_accuracy: 0.9654\n",
            "Epoch 78/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9210 - val_loss: 0.1046 - val_accuracy: 0.9148\n",
            "Epoch 79/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9199 - val_loss: 0.1046 - val_accuracy: 0.9097\n",
            "Epoch 80/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1040 - accuracy: 0.9190 - val_loss: 0.1046 - val_accuracy: 0.9011\n",
            "Epoch 81/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9192 - val_loss: 0.1046 - val_accuracy: 0.9179\n",
            "Epoch 82/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.9207 - val_loss: 0.1046 - val_accuracy: 0.9341\n",
            "Epoch 83/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9214 - val_loss: 0.1046 - val_accuracy: 0.9240\n",
            "Epoch 84/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9227 - val_loss: 0.1046 - val_accuracy: 0.9378\n",
            "Epoch 85/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.9238 - val_loss: 0.1046 - val_accuracy: 0.9449\n",
            "Epoch 86/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.9244 - val_loss: 0.1046 - val_accuracy: 0.9167\n",
            "Epoch 87/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9221 - val_loss: 0.1046 - val_accuracy: 0.9044\n",
            "Epoch 88/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9212 - val_loss: 0.1046 - val_accuracy: 0.8906\n",
            "Epoch 89/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9210 - val_loss: 0.1046 - val_accuracy: 0.9201\n",
            "Epoch 90/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9223 - val_loss: 0.1046 - val_accuracy: 0.9461\n",
            "Epoch 91/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9221 - val_loss: 0.1046 - val_accuracy: 0.9488\n",
            "Epoch 92/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.9232 - val_loss: 0.1046 - val_accuracy: 0.8920\n",
            "Epoch 93/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9221 - val_loss: 0.1046 - val_accuracy: 0.8819\n",
            "Epoch 94/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9195 - val_loss: 0.1046 - val_accuracy: 0.9473\n",
            "Epoch 95/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1041 - accuracy: 0.9203 - val_loss: 0.1046 - val_accuracy: 0.9172\n",
            "Epoch 96/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1046 - accuracy: 0.9234 - val_loss: 0.1046 - val_accuracy: 0.8926\n",
            "Epoch 97/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1043 - accuracy: 0.9232 - val_loss: 0.1046 - val_accuracy: 0.8981\n",
            "Epoch 98/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1045 - accuracy: 0.9211 - val_loss: 0.1046 - val_accuracy: 0.9458\n",
            "Epoch 99/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1042 - accuracy: 0.9226 - val_loss: 0.1046 - val_accuracy: 0.9545\n",
            "Epoch 100/100\n",
            "189/189 [==============================] - 1s 4ms/step - loss: 0.1044 - accuracy: 0.9218 - val_loss: 0.1046 - val_accuracy: 0.9123\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "mLGADfZb2RPo",
        "outputId": "09081ccd-ef6b-44e5-d730-a2f434d1e5a2"
      },
      "source": [
        "filepath = './models/ae_multi.json'\n",
        "weightspath = './weights/ae_multi.h5'\n",
        "if (not path.isfile(filepath)):  \n",
        "  # serialize model to JSON\n",
        "  autoencoder_json = autoencoder.to_json()\n",
        "  with open(filepath, \"w\") as json_file:\n",
        "    json_file.write(autoencoder_json)\n",
        "\n",
        "  # serialize weights to HDF5\n",
        "  autoencoder.save_weights(weightspath)\n",
        "  print(\"Saved model to disk\")\n",
        "\n",
        "  # load json and create model\n",
        "  json_file = open(filepath, 'r')\n",
        "  loaded_model_json = json_file.read()\n",
        "  json_file.close()\n",
        "  autoencoder = model_from_json(loaded_model_json)\n",
        "\n",
        "  # load weights into new model\n",
        "  autoencoder.load_weights(weightspath)\n",
        "  print(\"Loaded model from disk\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Saved model to disk\n",
            "Loaded model from disk\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yZKaeUx82Plg"
      },
      "source": [
        "# defining loss function, optimizer, metrics and then compiling model\n",
        "autoencoder.compile(optimizer='adam', loss='mean_squared_error',metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "k6FP2WyL45il",
        "outputId": "346f84fc-47a4-4f00-8eb2-8330b9ffc04b"
      },
      "source": [
        "# predicting target attribute on testing dataset\n",
        "test_results = autoencoder.evaluate(X_test, X_test, verbose=1)\n",
        "print(f'Test results - Loss: {test_results[0]} - Accuracy: {test_results[1]*100}%')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "985/985 [==============================] - 1s 989us/step - loss: 0.1044 - accuracy: 0.9095\n",
            "Test results - Loss: 0.10456941276788712 - Accuracy: 91.22689962387085%\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "ozvJupCe8nd_",
        "outputId": "fda15698-b278-4682-8374-8ff1f96fd3e8"
      },
      "source": [
        "# Plot of loss vs epoch of train and test dataset\n",
        "plt.plot(history['loss'])\n",
        "plt.plot(history['val_loss'])\n",
        "plt.title(\"Plot of loss vs epoch for train and test dataset\")\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='upper right')\n",
        "plt.savefig('plots/ae_multi_loss.png')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "cYRK9MR-8rPN",
        "outputId": "16619e31-bd72-4091-9a0a-9e67108ec9fe"
      },
      "source": [
        "# Plot of loss vs epoch of train and test dataset\n",
        "plt.plot(history['accuracy'])\n",
        "plt.plot(history['val_accuracy'])\n",
        "plt.title(\"Plot of accuracy vs epoch for train and test dataset\")\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='best')\n",
        "plt.savefig('plots/ae_multi_accuracy.png')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 312
        },
        "id": "B5sSYUlU5AQd",
        "outputId": "f1f5849a-db04-4d36-ff31-03a5c501e2da"
      },
      "source": [
        "# representation of model layers\n",
        "plot_model(autoencoder, to_file='plots/ae_multi.png', show_shapes=True,)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 216
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EL_GdCLX8xmB"
      },
      "source": [
        "# calculating reconstruction error\n",
        "predictions = autoencoder.predict(X_test)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HiA_107u8443"
      },
      "source": [
        "mse = np.mean(np.power(X_test - predictions, 2), axis=1)\n",
        "error_df = pd.DataFrame({'reconstruction_error': mse,\n",
        "                        'true_class': y_test})"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 297
        },
        "id": "7p0ih_628858",
        "outputId": "a4c5e3f3-3d2e-4d46-9e56-7b68538175a0"
      },
      "source": [
        "error_df.describe()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>reconstruction_error</th>\n",
              "      <th>true_class</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>count</th>\n",
              "      <td>31494.000000</td>\n",
              "      <td>31494.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>mean</th>\n",
              "      <td>0.104569</td>\n",
              "      <td>2.242840</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>std</th>\n",
              "      <td>0.043572</td>\n",
              "      <td>1.903732</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>min</th>\n",
              "      <td>0.040634</td>\n",
              "      <td>0.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25%</th>\n",
              "      <td>0.074501</td>\n",
              "      <td>0.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50%</th>\n",
              "      <td>0.083821</td>\n",
              "      <td>4.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75%</th>\n",
              "      <td>0.151404</td>\n",
              "      <td>4.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>max</th>\n",
              "      <td>0.229786</td>\n",
              "      <td>4.000000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "       reconstruction_error    true_class\n",
              "count          31494.000000  31494.000000\n",
              "mean               0.104569      2.242840\n",
              "std                0.043572      1.903732\n",
              "min                0.040634      0.000000\n",
              "25%                0.074501      0.000000\n",
              "50%                0.083821      4.000000\n",
              "75%                0.151404      4.000000\n",
              "max                0.229786      4.000000"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 219
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aLLK_7t_r1Ey"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}