2. Global  Explanation Methods.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "view-in-github"
   },
   "source": [
    "<a href=\"https://colab.research.google.com/github/danpele/2024-XAI-Intro/blob/main/2.%20Global%20%20Explanation%20Methods.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "grvYBSqLwztN"
   },
   "source": [
    "# Global Explanation Methods\n",
    "We start by creating the XGBoost-based credit risk model. XGBoost models are powerful but can be complex and difficult to interpret, making them an excellent example of a non-explainable model that we can diagnose and eventually interpret using the explanation methods discussed in the session.\n",
    "\n",
    "The outline of today's session is as follows:\n",
    "\n",
    " - Create black-box XGBoost classifier\n",
    " - Explain the black-box using a linear regression surrogate model\n",
    " - Compute permutation-based feature importance\n",
    " - Examine partial dependence"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "executionInfo": {
     "elapsed": 1444,
     "status": "ok",
     "timestamp": 1680274272279,
     "user": {
      "displayName": "Stefan Lessmann",
      "userId": "06342662613942148717"
     },
     "user_tz": -120
    },
    "id": "h6vhKogUqLQY"
   },
   "outputs": [],
   "source": [
    "# Standard libraries for data data handling and plotting\n",
    "import numpy as np \n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "from sklearn.preprocessing import MinMaxScaler"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "7ES6apIFFxqd"
   },
   "source": [
    "Below, we provide the code which you can run to load our credit data, train an XGBoost classifier, and produce test set predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 467
    },
    "executionInfo": {
     "elapsed": 8846,
     "status": "ok",
     "timestamp": 1680274306242,
     "user": {
      "displayName": "Stefan Lessmann",
      "userId": "06342662613942148717"
     },
     "user_tz": -120
    },
    "id": "rJcyttMeFsGC",
    "outputId": "1a263d66-0130-4723-c1d7-e6180688f519"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.93\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "\n",
    "# 1. Load data from GitHub\n",
    "data_url = \"https://raw.githubusercontent.com/danpele/2024-XAI-Intro/master/hmeq.csv\"\n",
    "df = pd.read_csv(data_url)\n",
    "\n",
    "# 2. Perform some minimal preprocessing to facilitate training a NN\n",
    "df = df.drop(columns=[\"REASON\", \"JOB\"], axis=1)  # delete categorical features\n",
    "y = df.pop(\"BAD\")  # Remove the target variable so that it remains unchanged\n",
    "df = df.fillna(value=df.mean())  # Mean-impute missing values\n",
    "scaler = MinMaxScaler(feature_range=(0,1))  # Scale feature values\n",
    "X = pd.DataFrame(scaler.fit_transform(df), columns=df.columns)\n",
    "\n",
    "# 3. Partition data into training (70%) and test (30%) sets\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# Split the data into training and testing sets\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)\n",
    "\n",
    "# 4. Train XGB\n",
    "import xgboost as xgb\n",
    "from sklearn.metrics import accuracy_score\n",
    "model = xgb.XGBClassifier(eval_metric='logloss')\n",
    "model.fit(X_train, y_train)\n",
    "\n",
    "# Evaluate the model\n",
    "\n",
    "yhat = model.predict(X_test)\n",
    "accuracy = accuracy_score(y_test, yhat)\n",
    "print(f\"Accuracy: {accuracy:.2f}\")\n",
    "\n",
    "\n",
    "# Compute a performance indicator over the test set predictions. \n",
    "\n",
    "from sklearn.metrics import ConfusionMatrixDisplay\n",
    "\n",
    "ConfusionMatrixDisplay.from_predictions(y_true=y_test, y_pred=yhat, labels=[0, 1], display_labels=['Good(0)', 'Bad(1)']);\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9X_bAobqH9qI"
   },
   "source": [
    "## Surrogate model-based explanation\n",
    "A surrogate model is, by definition, an intrinsically interpretable model, which we train to predict (aka reproduce) the predictions of a black-box model.\n",
    "\n",
    "We just created a black-box model above. As surrogate model, we will use linear regression, and see how well we can linearly approximate our XGB. Specifically, fit the surrogate model to the predicted class *probabilities* of the neural network using the original features. To predict probabilities, you will need to use the method `predict_proba()`, which any `sklearn` classifier provides.\n",
    "\n",
    "One question worth discussing in this context is which data set to use for fitting the surrogate model, the training or test data. From a programming perspective, it does not make a difference. So first develop your code and then we discuss the subtle differences between the two options."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 449
    },
    "executionInfo": {
     "elapsed": 727,
     "status": "ok",
     "timestamp": 1680274413673,
     "user": {
      "displayName": "Stefan Lessmann",
      "userId": "06342662613942148717"
     },
     "user_tz": -120
    },
    "id": "uBUZ3_vLH3Sp",
    "outputId": "64294999-4a7c-43ff-e92b-87d06c3c1c22"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlMAAAGwCAYAAACNeeBZAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8fJSN1AAAACXBIWXMAAA9hAAAPYQGoP6dpAABHGUlEQVR4nO3deVxV1f7/8fcG5CAiRxwhQ3GGNOe5yVlzuFbmlBNlWjfTLMupm2MKmTaaWl0Bu6V2UyvLIefym0OWkhYoZpKW0ODAMU0MWL8/+nluJ9DQDTL4ej4e+/For7322p+9Oo94t/bmYBljjAAAAHBFvAq6AAAAgKKMMAUAAGADYQoAAMAGwhQAAIANhCkAAAAbCFMAAAA2EKYAAABs8CnoAq4FWVlZOnbsmEqXLi3Lsgq6HAAAkAvGGJ0+fVrXXXedvLwuvv5EmLoKjh07ptDQ0IIuAwAAXIGjR4/q+uuvv+hxwtRVULp0aUl//MsIDAws4GoAAEBuuFwuhYaGun+OXwxh6iq48GgvMDCQMAUAQBHzd6/o8AI6AACADYQpAAAAGwhTAAAANhCmAAAAbCBMAQAA2ECYAgAAsIGvRgBw1YWNX1XQJQDFXnJ0t4Iu4ZrByhQAAIANhCkAAAAbCFMAAAA25HmYioyMlGVZsixLJUqUUKVKldSxY0fFxMQoKyvL3S8sLMzd789bdHS0JCk5Odmj3dfXVzVr1tTTTz8tY8wlx7iwtWnTxt3vhRdeyHbtHTt2eNQ+evRo9zkXuFwuPfnkkwoPD5efn5+Cg4PVoUMHrVixwl0HAAC4duXLC+hdunRRbGysMjMz9eOPP2rt2rV65JFHtGzZMq1cuVI+Pn9cdtq0aRo2bJjHuX/9Y4IbNmxQ3bp1lZ6erv/7v//T/fffr5CQEA0dOlS7du1SZmamJGnbtm3q1auXDhw44P77d76+vhet0c/PT+PGjdPHH3980T6nTp3SzTffrLS0ND399NNq1qyZfHx89PHHH2vs2LFq166dypQpcyVTBAAAiol8CVMOh0PBwcGSpMqVK6tx48Zq2bKl2rdvr7i4ON1///2S/ghOF/pdTLly5dx9qlatqpiYGO3evVtDhw5VhQoV3P3Kli0rSapYsWKuAs4DDzyg+fPna/Xq1eratWuOfSZOnKjk5GQlJSXpuuuuc7fXrl1b/fv3l5+f399eBwAAFG9X7Z2pdu3aqUGDBlqxYsUVj/H5559r9+7datGihe16wsLC9OCDD2rChAkejx8vyMrK0tKlSzVgwACPIHVBQECAe4Xtr9LT0+VyuTw2AABQPF3VF9DDw8OVnJzs3h83bpwCAgI8ti1btnic07p1awUEBMjX11fNmjVTnz59NHjw4Dyp51//+pcOHz6st956K9uxX375RSdPnlR4ePhljxsVFSWn0+neQkND86JcAABQCF3VL+00xsiyLPf+E088ocjISI8+lStX9th/++23FRERod9//1379u3TqFGjFBQU5H5R3Y4KFSro8ccf16RJk9S3b99stUryqDe3JkyYoMcee8y973K5CFQAABRTVzVMJSYmqlq1au798uXLq2bNmpc8JzQ01N0nIiJC3377rZ566ilNmTIlT95ZeuyxxzRv3jzNmzfPo71ChQoKCgpSYmLiZY/pcDjkcDhs1wYAAAq/q/aYb9OmTdq3b5969eplaxxvb29lZGTo/PnzeVJXQECAnnrqKc2YMcPj3SYvLy/17dtXb731lo4dO5btvDNnzigjIyNPagAAAEVXvoSp9PR0paam6ocfftDu3bs1c+ZM9ezZU927d/d43+n06dNKTU312P76svbx48eVmpqq77//XmvWrNGLL76otm3bur/+IC8MHz5cTqdTS5Ys8WifOXOmQkND1aJFC73xxhtKSEjQwYMHFRMTo4YNG+rXX3/NsxoAAEDRlC+P+dauXauQkBD5+PgoKChIDRo00EsvvaQhQ4bIy+t/+W3SpEmaNGmSx7kPPPCAFixY4N7v0KGDpD9WpEJCQtS1a1fNmDEjT+stUaKEpk+frnvuucejPSgoSDt27FB0dLSefvppfffddwoKCtKNN96oZ599Vk6nM0/rAAAARY9l+BrvfOdyueR0OpWWlpanK2pAURU2flVBlwAUe8nR3Qq6hCIvtz+/+dt8AAAANlzV3+YDAIn/YwZQvLAyBQAAYANhCgAAwAbCFAAAgA2EKQAAABsIUwAAADYQpgAAAGwgTAEAANhAmAIAALCBMAUAAGADYQoAAMAGwhQAAIANhCkAAAAbCFMAAAA2EKYAAABsIEwBAADYQJgCAACwwaegCwBQvISNX/W3fZKju12FSgDg6mBlCgAAwAbCFAAAgA2EKQAAABsIUwAAADYUuTCVmpqqkSNHqnr16nI4HAoNDVWPHj20ceNGSVJYWJheeOGFHM9NTk6WZVmqWLGiTp8+7XGsYcOGmjJlikfb119/rT59+qhChQpyOByqVauWnnrqKZ09ezY/bg0AABRBRSpMJScnq0mTJtq0aZNmzZqlffv2ae3atWrbtq1GjBiR63FOnz6t2bNnX7LPjh071KJFC50/f16rVq1SUlKSZs6cqUWLFqljx446f/683dsBAADFQJH6aoSHHnpIlmXps88+U6lSpdztdevW1X333ZfrcUaOHKnnnntOI0aMUMWKFbMdN8Zo6NChioiI0IoVK+Tl9UfmrFq1qmrXrq1GjRrp+eef17hx4+zfFAAAKNKKzMrUiRMntHbtWo0YMcIjSF1QpkyZXI/Vv39/1axZU9OmTcvxeHx8vBISEvTYY4+5g9QFDRo0UIcOHbRkyZKLjp+eni6Xy+WxAQCA4qnIhKlvvvlGxhiFh4fbHsuyLEVHR+u1117ToUOHsh1PSkqSJEVEROR4fkREhLtPTqKiouR0Ot1baGio7ZoBAEDhVGTClDFG0h9BKC907txZN998s5566qkrquVSdUyYMEFpaWnu7ejRo3ZKBQAAhViRCVO1atWSZVlKTEzMszGjo6P19ttva8+ePR7ttWvXliQlJCTkeN7+/ftVq1ati47rcDgUGBjosQEAgOKpyISpsmXLqnPnznrllVd05syZbMdPnTp12WM2b95cd911l8aPH+/R3rBhQ4WHh+v5559XVlaWx7Evv/xSGzZsUP/+/S/7egAAoPgpMmFKkubNm6fMzEw1b95cy5cv18GDB5WYmKiXXnpJrVq1cvf74YcfFB8f77GdOHEixzFnzJihTZs26cCBA+42y7L073//WwkJCerVq5c+++wzHTlyRO+884569OihVq1aafTo0fl9uwAAoAgoUmGqWrVq2r17t9q2basxY8aoXr166tixozZu3Kj58+e7+82ePVuNGjXy2FauXJnjmLVr19Z9992nc+fOebTfdNNN2rFjh7y9vdW1a1fVrFlTEyZM0JAhQ7R+/Xo5HI58vVcAAFA0WObCm93INy6XS06nU2lpabw/hWIvbPyqv+2THN3tKlQCAPbk9ud3kVqZAgAAKGyK1DegAyj8WHUCcK1hZQoAAMAGwhQAAIANhCkAAAAbCFMAAAA2EKYAAABsIEwBAADYQJgCAACwgTAFAABgA2EKAADABsIUAACADYQpAAAAGwhTAAAANhCmAAAAbCBMAQAA2ECYAgAAsIEwBQAAYINPQRcA4NoTNn5VQZeQZ5KjuxV0CQAKGCtTAAAANhCmAAAAbCBMAQAA2ECYAgAAsKHIhqnIyEhZlqXo6GiP9vfee0+WZUmStmzZIsuydOrUKY/9evXqKTMz0+O8MmXKKC4uzqNt27Zt6tq1q4KCguTn56cbb7xRc+bMyXYuAAC4dhXZMCVJfn5+euaZZ3Ty5MnLOu/QoUN64403Ltnn3Xff1W233abrr79emzdv1v79+/XII49oxowZ6tevn4wxdkoHAADFRJEOUx06dFBwcLCioqIu67yRI0dq8uTJOnfuXI7Hz5w5o2HDhukf//iHXnvtNTVs2FBhYWG6//77tWjRIi1btkz//e9/Lzp+enq6XC6XxwYAAIqnIh2mvL29NXPmTL388sv6/vvvc33e6NGjlZGRoblz5+Z4fN26dTp+/Lgef/zxbMd69Oih2rVra8mSJRcdPyoqSk6n072FhobmujYAAFC0FOkwJUl33nmnGjZsqMmTJ+f6HH9/f02ePFlRUVFKS0vLdjwpKUmSFBERkeP54eHh7j45mTBhgtLS0tzb0aNHc10bAAAoWop8mJKkZ555RosWLVJCQkKuzxk6dKjKly+vZ5555qJ9LvZelDFGvr6+Fz3P4XAoMDDQYwMAAMVTsQhTt956qzp37qyJEyfm+hwfHx89/fTTevHFF3Xs2DGPY7Vq1ZIkJSYm5nju/v37Vbt27SsvGAAAFBvFIkxJUnR0tD744ANt27Yt1+f07t1bdevW1dSpUz3aO3furLJly2rOnDnZzlm5cqUOHjyoyMhIuyUDAIBioNiEqRtvvFEDBgzQyy+/fFnnRUdHKyYmRmfOnHG3lSpVSq+++qref/99DR8+XHv37lVycrIWLlyoyMhI3X///eratWte3wIAACiCik2YkqTp06df9vc/tWvXTu3atVNGRoZH+913363NmzfryJEjuuWWW1StWjXdf//9GjdunF5//fW8LBsAABRhluHbJ3Pl3Llz6tmzp44ePaqPP/5YFSpUyPW5LpdLTqdTaWlpvIwOSAobv6qgS8gzydHdCroEAPkktz+/i9XKVH7y8/PT+++/r8GDB+uTTz4p6HIAAEAhwcrUVcDKFAAARQ8rUwAAAFcBYQoAAMAGwhQAAIANhCkAAAAbCFMAAAA2EKYAAABsIEwBAADYQJgCAACwgTAFAABgA2EKAADABsIUAACADYQpAAAAGwhTAAAANhCmAAAAbCBMAQAA2OBT0AXAnrDxqwq6BOCyJUd3K+gSACDPsDIFAABgA2EKAADABsIUAACADYQpAAAAGwpdmEpNTdXIkSNVvXp1ORwOhYaGqkePHtq4caMkKSwsTC+88MLfjrN48WJ5e3vrwQcfzPG4y+XSU089pbp166pkyZIqV66cmjVrplmzZunkyZPufm3atJFlWdm2i40LAACuLYXqt/mSk5N10003qUyZMpo1a5bq16+v33//XR999JFGjBih/fv353qsmJgYjR07VvPnz9dzzz0nf39/97ETJ07o5ptvlsvl0vTp09WkSRP5+vrqm2++0eLFi7V48WKNGDHC3X/YsGGaNm2ax/h/Hg8AAFy7ClWYeuihh2RZlj777DOVKlXK3V63bl3dd999uR4nOTlZ27Zt0/Lly7V582YtW7ZMgwcPdh+fOHGijhw5ogMHDqhy5cru9vDwcHXv3l3GGI/x/P39FRwcbOPOAABAcVVoHvOdOHFCa9eu1YgRIzyC1AVlypTJ9VgxMTHq1q2bnE6nBg4cqIULF7qPZWVl6e2339bAgQM9gtSfWZZ12fX/WXp6ulwul8cGAACKp0ITpr755hsZYxQeHm5rnKysLMXFxWngwIGSpH79+mn79u365ptvJEk///yzTp06pTp16nic16RJEwUEBCggIED9+/f3ODZv3jz3sQvbokWLLlpDVFSUnE6newsNDbV1TwAAoPAqNGHqwqM1u6tC69at05kzZ3T77bdLksqXL69OnTopJibGo99fr/Puu+8qPj5enTt31m+//eZxbMCAAYqPj/fY7rzzzovWMGHCBKWlpbm3o0eP2ronAABQeBWad6Zq1aoly7KUmJioO+6444rHiYmJ0YkTJzxeEM/KytKePXs0ffp0VahQQWXKlMn2MnuVKlUkSaVLl9apU6c8jjmdTtWsWTPXNTgcDjkcjiu+BwAAUHQUmpWpsmXLqnPnznrllVd05syZbMf/GnBycvz4cb3//vtaunRptpWkX3/9VWvWrJGXl5f69OmjN998Uz/88EM+3AkAALiWFJqVKemPd5Nat26t5s2ba9q0aapfv74yMjK0fv16zZ8/X4mJiZKkH374QfHx8R7nVqlSRf/5z39Urlw59e7dW15enjmxe/fuWrhwobp3766ZM2dqy5YtatGihaZNm6amTZuqVKlS2rt3r7Zv36569ep5nHv27FmlpqZ6tDkcDgUFBeX9JAAAgCKlUIWpatWqaffu3ZoxY4bGjBmjlJQUVahQQU2aNNH8+fPd/WbPnq3Zs2d7nBsbG6uYmBjdeeed2YKUJPXq1Ut9+/bVjz/+qEqVKumzzz7TM888o2effVaHDx+Wl5eXatWqpb59+2r06NEe577++ut6/fXXPdo6d+6stWvX5t3NAwCAIskyf/1SJeQ5l8slp9OptLQ0BQYG5unYYeNX5el4wNWQHN2toEsAgL+V25/fheadKQAAgKKoUD3mw+Xj//ABAChYrEwBAADYQJgCAACwgTAFAABgA2EKAADABsIUAACADYQpAAAAGwhTAAAANhCmAAAAbCBMAQAA2ECYAgAAsIEwBQAAYANhCgAAwAbCFAAAgA2EKQAAABsIUwAAADYQpgAAAGzwKegCAFx7wsavKugSABQjydHdCvT6rEwBAADYQJgCAACwgTAFAABgA2EKAADAhgIJU5GRkbIsS5ZlqUSJEqpUqZI6duyomJgYZWVlufuFhYW5+/15i46OliQlJyfLsizFx8fneJ24uDiVKVPGY9+yLHXp0sWj36lTp2RZlrZs2eLR/uGHH6pNmzYqXbq0/P391axZM8XFxeXFFAAAgGKiwFamunTpopSUFCUnJ2vNmjVq27atHnnkEXXv3l0ZGRnuftOmTVNKSorHNnLkyCu+ro+PjzZu3KjNmzdfst/LL7+snj17qnXr1tq5c6f27t2rfv366cEHH9Tjjz9+xdcHAADFS4F9NYLD4VBwcLAkqXLlymrcuLFatmyp9u3bKy4uTvfff78kqXTp0u5+eaFUqVLq06ePxo8fr507d+bY5+jRoxozZoxGjx6tmTNnutvHjBkjX19fjRo1Sr1791aLFi1yPD89PV3p6enufZfLlWf1AwCAwqVQvTPVrl07NWjQQCtWrMjX60yZMkX79u3TsmXLcjy+bNky/f777zmuQD3wwAMKCAjQkiVLLjp+VFSUnE6newsNDc2z2gEAQOFSqMKUJIWHhys5Odm9P27cOAUEBHhsf3236XJdd911euSRR/Tkk096PFK8ICkpSU6nUyEhIdmO+fr6qnr16kpKSrro+BMmTFBaWpp7O3r0qK16AQBA4VXovgHdGCPLstz7TzzxhCIjIz36VK5c2fZ1xo0bp1dffVUxMTHq06fPZdfo6+t70eMOh0MOh8NuiQAAoAgodCtTiYmJqlatmnu/fPnyqlmzpsdWsmRJ29cpU6aMJkyYoKlTp+rs2bMex2rVqqW0tDQdO3Ys23nnz5/Xt99+q9q1a9uuAQAAFH2FKkxt2rRJ+/btU69eva7K9UaOHCkvLy+9+OKLHu133323fHx8NGfOnGznLFiwQGfPntXgwYOvSo0AAKBwK7DHfOnp6UpNTVVmZqZ+/PFHrV27VlFRUerevbtHUDl9+rRSU1M9zvX391dgYKB7/8CBA9nGv+GGG/62Bj8/P02dOlUjRozwaK9SpYpmzZqlxx9/XH5+fho0aJBKlCih999/XxMnTtTTTz+tevXqXe4tAwCAYqjAwtTatWsVEhIiHx8fBQUFqUGDBnrppZc0ZMgQeXn9b8Fs0qRJmjRpkse5DzzwgBYsWODe79evX7bxDx8+nKs6hgwZojlz5ighIcGj/dFHH1X16tU1Z84cvfjiizpz5owkacmSJTleDwAAXJssY4wp6CKKghMnTqh9+/YKDAzUmjVr5O/vn+tzXS6XnE6n0tLSPFbUgGtV2PhVBV0CgGIkObpbvoyb25/fheqdqcKsbNmy2rBhg9q3b6/t27cXdDkAAKCQYGXqKmBlCgCAooeVKQAAgKuAMAUAAGADYQoAAMAGwhQAAIANhCkAAAAbCFMAAAA2EKYAAABsIEwBAADYQJgCAACwgTAFAABgA2EKAADABsIUAACADYQpAAAAGwhTAAAANhCmAAAAbCBMAQAA2OBT0AUAuPaEjV9V0CVcE5KjuxV0CcA1gZUpAAAAGwhTAAAANhCmAAAAbCiwMBUZGSnLsmRZlkqUKKFKlSqpY8eOiomJUVZWlrtfWFiYu9+ft+joaElScnKyR7vT6VTLli31wQcfZLvmb7/9psmTJ6tOnTpyOBwqX7687r77bn399dfZ+rpcLj311FOqW7euSpYsqXLlyqlZs2aaNWuWTp48mX8TAwAAipQCXZnq0qWLUlJSlJycrDVr1qht27Z65JFH1L17d2VkZLj7TZs2TSkpKR7byJEjPcbasGGDUlJStHPnTjVv3ly9evXSV1995T6enp6uDh06KCYmRtOnT1dSUpJWr16tzMxMtWjRQjt27HD3PXHihFq2bKnY2Fg9/vjj2rlzpz799FNNnjxZ8fHxWrx4cf5PDgAAKBIK9Lf5HA6HgoODJUmVK1dW48aN1bJlS7Vv315xcXG6//77JUmlS5d297uYcuXKKTg4WMHBwZoxY4Zefvllbd68WfXq1ZMkvfDCC9q+fbv27NmjBg0aSJKqVq2q5cuXq0WLFho6dKi++uorWZaliRMn6siRIzpw4IAqV67svkZ4eLi6d+8uY0x+TAcAACiCCt07U+3atVODBg20YsWKKzr/999/1+uvvy5JKlGihLt98eLF6tixoztIXeDl5aVHH31UCQkJ+vLLL5WVlaW3335bAwcO9AhSf2ZZ1iVrSE9Pl8vl8tgAAEDxVOjClPTHClBycrJ7f9y4cQoICPDYtmzZ4nFO69atFRAQID8/P40ZM0ZhYWHq06eP+3hSUpIiIiJyvN6F9qSkJP388886deqU6tSp49GnSZMm7mv379//kvVHRUXJ6XS6t9DQ0Mu4ewAAUJQUyi/tNMZ4rP488cQTioyM9Ojz11Wjt99+W+Hh4UpKStLo0aO1YMEClS1bNtfXkzxXnP66+vTuu+/q/PnzGjdunH777bdLjjdhwgQ99thj7n2Xy0WgAgCgmCqUYSoxMVHVqlVz75cvX141a9a85DmhoaGqVauWatWqpYCAAPXq1UsJCQmqWLGiJKl27dpKSEjI8dz9+/dLkmrVqqUKFSqoTJky7rYLqlSpIumP97dOnTp1yVocDoccDscl+wAAgOKh0D3m27Rpk/bt26devXpd8Ri33Xab6tWrpxkzZrjb+vXrpw0bNujLL7/06JuVlaXnn39eN9xwgxo0aCAvLy/16dNHb775pn744YcrrgEAAFwbCjRMpaenKzU1VT/88IN2796tmTNnqmfPnurevbsGDx7s7nf69GmlpqZ6bH/3UveYMWP06quvugPRo48+qubNm6tHjx565513dOTIEe3atUu9evVSYmKiFi5c6H60N3PmTFWuXFktWrRQTEyM9u7dq0OHDundd9/V9u3b5e3tnX+TAgAAipQCDVNr165VSEiIwsLC1KVLF23evFkvvfSS3n//fY/AMmnSJIWEhHhsY8eOveTY3bt3V1hYmHt1ys/PT5s2bdKQIUM0ceJE1axZU126dJG3t7d27Nihli1bus8tV66cPvvsMw0ePFjPPvusmjdvrhtvvFFTpkxR37593b8tCAAAYBm+NCnfuVwuOZ1OpaWlKTAwsKDLAQpc2PhVBV3CNSE5ultBlwAUabn9+V3o3pkCAAAoSgrlb/MBKN5YMQFQnLAyBQAAYANhCgAAwAbCFAAAgA2EKQAAABsIUwAAADYQpgAAAGwgTAEAANhAmAIAALCBMAUAAGADYQoAAMAGwhQAAIANhCkAAAAbCFMAAAA2EKYAAABsIEwBAADYQJgCAACwgTAF4KoKG7+qoEsAgDxFmAIAALCBMAUAAGADYQoAAMAGwhQAAIAN10SYMsaoQ4cO6ty5c7Zj8+bNk9Pp1JEjR/Thhx+qTZs2Kl26tPz9/dWsWTPFxcV59E9OTpZlWYqPj786xQMAgELtmghTlmUpNjZWO3fu1KuvvupuP3z4sMaNG6cXX3xR77//vnr27KnWrVtr586d2rt3r/r166cHH3xQjz/+eAFWDwAACjOfgi7gagkNDdWLL76ohx9+WJ06dVJYWJiGDh2q9u3bq3379qpRo4ZGjx6tmTNnus8ZM2aMfH19NWrUKPXu3VstWrQowDsAAACF0TWxMnXBkCFD1L59e917772aO3euvvrqK7322mtatmyZfv/99xxXoB544AEFBARoyZIlub5Oenq6XC6XxwYAAIqna2Zl6oLXXntN9erV09atW7Vs2TJVrFhRSUlJcjqdCgkJydbf19dX1atXV1JSUq6vERUVpalTp+Zl2QAAoJC6plamJKlixYoaPny4IiIidOedd+bqHGOMLMvK9TUmTJigtLQ093b06NErLRcAABRy11yYkiQfHx/5+PxvUa527dpKS0vTsWPHsvU9f/68vv32W9WqVSvX4zscDgUGBnpsAACgeLomw9Rf9erVSz4+PpozZ062YwsWLNCZM2fUv3//AqgMAAAUdtfcO1M5qVKlimbNmqXHH39cfn5+GjRokEqUKKH3339fEydO1JgxY/hNPgAAkCPC1P/36KOPqkaNGpo9e7ZefPFFZWZmqm7dupo/f77uvfded7+srCxJ8nhMCAAArl2WMcYUdBFFyY4dO9SqVSv9/PPPKl++fK7OcblccjqdSktL4/0pXPPCxq9ScnS3gi4DAP5Wbn9+s7ySSxkZGUpOTtazzz6rBg0a5DpIAQCA4o0X0HPpq6++Uv369ZWSkqI33nijoMsBiixWpQAUN6xM5VLDhg119uzZgi4DAAAUMqxMAQAA2ECYAgAAsIEwBQAAYANhCgAAwAbCFAAAgA2EKQAAABsIUwAAADYQpgAAAGwgTAEAANhAmAIAALCBMAUAAGADYQoAAMAGwhQAAIANhCkAAAAbCFMAAAA2+BR0AQCuPWHjVxV0Cde85OhuBV0CUGywMgUAAGADYQoAAMAGwhQAAIANhCkAAAAbCl2Y6tGjhzp06JDjse3bt8uyLO3evVuSNHz4cHl7e2vp0qXZ+k6ZMkUNGzbMcZzIyEjdcccd2drj4+NlWZaSk5MlSVu2bJFlWTluqampV3R/AACgeCl0YWro0KHatGmTvvvuu2zHYmJi1LBhQzVu3Fhnz57V22+/rSeeeEILFy7M15oOHDiglJQUj61ixYr5ek0AAFA0FLow1b17d1WsWFFxcXEe7RfC09ChQyVJ77zzjm644QZNmDBBn376qXs1KT9UrFhRwcHBHpuX18WnLj09XS6Xy2MDAADFU6ELUz4+Pho8eLDi4uJkjHG3v/POOzp//rwGDBggSVq4cKEGDhwop9Oprl27KjY2tqBKziYqKkpOp9O9hYaGFnRJAAAgnxS6MCVJ9913n5KTk7VlyxZ3W0xMjO666y4FBQXp4MGD2rFjh/r27StJGjhwoGJjY5WVlZUv9Vx//fUKCAhwb3Xq1Llk/wkTJigtLc29HT16NF/qAgAABa9QfgN6eHi4WrdurZiYGLVt21aHDh3S1q1btW7dOkl/rEp17txZ5cuXlyR17dpVQ4cO1YYNG9SpU6c8r2fr1q0qXbq0e9/H59LT5nA45HA48rwOAABQ+BTKlSnpjxfRly9fLpfLpdjYWFWtWlXt27dXZmam3njjDa1atUo+Pj7y8fGRv7+/Tpw4kesX0QMDA5WWlpat/dSpU5Ikp9Pp0V6tWjXVrFnTvYWFhdm9PQAAUEwU2jDVp08feXt7a/HixVq0aJHuvfdeWZal1atX6/Tp09qzZ4/i4+Pd2zvvvKP33ntPx48f/9uxw8PD9dVXX+ncuXMe7bt27VKFChUUFBSUX7cFAACKmUL5mE+SAgIC1LdvX02cOFFpaWmKjIyU9Mcjvm7duqlBgwYe/evWravRo0frzTff1COPPCJJ+u233xQfH59t3AEDBmj69OkaNGiQxo0bp6CgIG3fvl1RUVGaMGFCtlp++umnbMGrXLlyKlGiRN7dMAAAKJIK7cqU9MejvpMnT6pDhw6qUqWKfvzxR61atUq9evXK1teyLN11110ej/qSkpLUqFEjj+3++++X0+nU1q1bZYzRHXfcoQYNGmjWrFmaPn26xowZk23sOnXqKCQkxGP74osv8vXeAQBA0WCZP3//APKFy+WS0+lUWlqaAgMDC7ocoMCFjV9V0CVc85KjuxV0CUChl9uf34V6ZQoAAKCwK7TvTAEovlgVAVCcsDIFAABgA2EKAADABsIUAACADYQpAAAAGwhTAAAANhCmAAAAbCBMAQAA2ECYAgAAsIEwBQAAYANhCgAAwAbCFAAAgA2EKQAAABsIUwAAADYQpgAAAGwgTAEAANjgU9AFAJDCxq8q6BKuquTobgVdAgDkGVamAAAAbCBMAQAA2ECYAgAAsIEwBQAAYMNlhanIyEhZlqUHH3ww27GHHnpIlmUpMjLS3Xb06FENHTpU1113nXx9fVW1alU98sgjOn78uMe5bdq0kWVZsixLvr6+qlGjhiZMmKD09HTFxcW5j11s27Jli0c/b29vBQUFqUWLFpo2bZrS0tKyXW/06NHZ7uG9996TZVnu/Ytd28/P73KmDQAAFGOXvTIVGhqqpUuX6rfffnO3nTt3TkuWLFGVKlXcbd9++62aNm2qpKQkLVmyRN98840WLFigjRs3qlWrVjpx4oTHuMOGDVNKSoq++eYbzZo1S6+88oqmTJmivn37KiUlxb21atXK3ffC1rp1a0lSYGCgUlJS9P3332vbtm0aPny43njjDTVs2FDHjh27ogm6MOaft+++++6KxgIAAMXPZX81QuPGjfXtt99qxYoVGjBggCRpxYoVCg0NVfXq1d39RowYIV9fX61bt04lS5aUJFWpUkWNGjVSjRo19OSTT2r+/Pnu/v7+/goODnb3W7x4sdatW6eoqCj3+ZLk6+vr0ffPLMtyt4eEhCgiIkI9evRQ3bp1NXbsWL355puXe7seYwIAAPzVFb0zde+99yo2Nta9HxMTo/vuu8+9f+LECX300Ud66KGHPIKQJAUHB2vAgAF6++23ZYzJcfwvv/xSn376qUqUKHEl5XmoWLGiBgwYoJUrVyozM9P2eLmRnp4ul8vlsQEAgOLpisLUoEGD9H//939KTk7Wd999p08//VQDBw50Hz948KCMMYqIiMjx/IiICJ08eVI///yzu23evHkKCAiQw+FQw4YN9fPPP+uJJ564kvKyCQ8P1+nTp7O9q5UbaWlpCggI8Ng6dep0yXOioqLkdDrdW2ho6JWWDgAACrkr+gb08uXLq1u3blq0aJGMMerWrZvKly+f6/MvrEj9+WXvAQMG6Mknn5TL5dIzzzyjwMBA9erV60rKy9X1cqt06dLavXu3R9tfV9v+asKECXrsscfc+y6Xi0AFAEAxdcV/Tua+++7Tww8/LEl65ZVXPI7VrFlTlmUpISFBd9xxR7Zz9+/fr6CgII8A5nQ6VbNmTUnSm2++qbp162rhwoUaOnTolZbolpiYqMDAQJUrV07SHy+V//U3/CTp1KlTCgwM9Gjz8vJy15VbDodDDofjygsGAABFxhV/z1SXLl10/vx5nT9/Xp07d/Y4Vq5cOXXs2FHz5s3z+K0/SUpNTdVbb72lvn37XnSlqESJEpo4caL+9a9/6ezZs1daoiTpp59+0uLFi3XHHXfIy+uP2w0PD9fnn3+ere+uXbtUp04dW9cDAADXlisOU97e3kpMTFRiYqK8vb2zHZ87d67S09PVuXNnffLJJzp69KjWrl2rjh07qnLlypoxY8Ylx7/nnntkWZbmzZuX65qMMUpNTVVKSooSExMVExOj1q1by+l0Kjo62t3voYce0qFDhzRixAh9+eWXSkpK0iuvvKKFCxdme0/rwph/3bKysnJdFwAAKL5sfQN6YGBgtsdiF9SqVUuff/65atSoob59+6pGjRoaPny42rZtq+3bt6ts2bKXHNvX11cPP/ywZs2apV9//TVX9bhcLoWEhKhy5cpq1aqVXn31VQ0ZMkR79uxRSEiIu19YWJi2bt2qQ4cOqVOnTmrWrJni4uIUFxen3r175zjmX7effvopVzUBAIDizTIX+34C5BmXyyWn06m0tLSLhk9c28LGryroEq6q5OhuBV0CAPyt3P785m/zAQAA2HDFv80HIO+wUgMARRcrUwAAADYQpgAAAGwgTAEAANhAmAIAALCBMAUAAGADYQoAAMAGwhQAAIANhCkAAAAbCFMAAAA2EKYAAABsIEwBAADYQJgCAACwgTAFAABgA2EKAADABsIUAACADYQpAAAAG3wKugAA+Sds/KqCLiFHydHdCroEAMgzrEwBAADYQJgCAACwgTAFAABgA2EKAADAhiIXpiIjI3XHHXfkeOy3337T5MmTVadOHTkcDpUvX1533323vv766xz7L168WN7e3nrwwQezHduyZYssy1K9evWUmZnpcaxMmTKKi4uzeysAAKAYKHJh6mLS09PVoUMHxcTEaPr06UpKStLq1auVmZmpFi1aaMeOHdnOiYmJ0dixY7V06VKdPXs2x3EPHTqkN954I7/LBwAARVSxCVMvvPCCtm/frg8//FB9+vRR1apV1bx5cy1fvlwREREaOnSojDHu/snJydq2bZvGjx+v8PBwLVu2LMdxR44cqcmTJ+vcuXO5riU9PV0ul8tjAwAAxVOxCVOLFy9Wx44d1aBBA492Ly8vPfroo0pISNCXX37pbo+JiVG3bt3kdDo1cOBALVy4MMdxR48erYyMDM2dOzfXtURFRcnpdLq30NDQK7spAABQ6BWbMJWUlKSIiIgcj11oT0pKkiRlZWUpLi5OAwcOlCT169dP27dv1zfffJPtXH9/f02ePFlRUVFKS0vLVS0TJkxQWlqaezt69OiV3BIAACgCik2YupQLj/csy5IkrVu3TmfOnNHtt98uSSpfvrw6deqkmJiYHM8fOnSoypcvr2eeeSZX13M4HAoMDPTYAABA8VRswlTt2rWVkJCQ47H9+/dLkmrVqiXpj0d8J06ckL+/v3x8fOTj46PVq1dr0aJF2X5zT5J8fHz09NNP68UXX9SxY8fy7yYAAECRU2zCVL9+/bRhwwaP96KkPx7pPf/887rhhhvUoEEDHT9+XO+//76WLl2q+Ph4j+3XX3/VmjVrchy/d+/eqlu3rqZOnXo1bgcAABQRRfIPHaelpSk+Pt6jbcCAAXr//ffVo0cPzZkzRy1atNCPP/6omTNnKjExURs2bJBlWfrPf/6jcuXKqXfv3vLy8syS3bt318KFC9W9e/ccrxsdHa3OnTvn120BAIAiqEiGqS1btqhRo0YebUOGDNGmTZsUFRWliRMn6rvvvlPp0qXVtm1b7dixQ/Xq1ZP0xyO+O++8M1uQkqRevXqpb9+++vHHH3O8brt27dSuXTutW7cu728KAAAUSZb585cvIV+4XC45nU6lpaXxMjquqrDxqwq6hBwlR3cr6BIA4G/l9ud3sXlnCgAAoCAUycd8AHKHFSAAyH+sTAEAANhAmAIAALCBMAUAAGADYQoAAMAGwhQAAIANhCkAAAAbCFMAAAA2EKYAAABsIEwBAADYQJgCAACwgTAFAABgA2EKAADABsIUAACADYQpAAAAGwhTAAAANhCmAAAAbPAp6AIAXHvCxq/Kk3GSo7vlyTgAYAcrUwAAADYQpgAAAGwgTAEAANiQp2EqMjJSd9xxR677W5al9957Ly9LyFdTpkxRw4YNC7oMAABQiBSLlanff/+9oEsAAADXqHwLU23atNGoUaM0duxYlS1bVsHBwZoyZYr7eFhYmCTpzjvvlGVZ7n1J+uCDD9SkSRP5+fmpevXqmjp1qjIyMtzHLcvSggUL1LNnT5UqVUpPP/20JGnlypVq2rSp/Pz8VL58ed11113uc86fP6+xY8eqcuXKKlWqlFq0aKEtW7a4j8fFxalMmTJ67733VLt2bfn5+aljx446evSo+/jUqVP15ZdfyrIsWZaluLi4PJ83AABQtOTrytSiRYtUqlQp7dy5U7NmzdK0adO0fv16SdKuXbskSbGxsUpJSXHvf/TRRxo4cKBGjRqlhIQEvfrqq4qLi9OMGTM8xp48ebJ69uypffv26b777tOqVat01113qVu3btqzZ482btyopk2buvvfe++9+vTTT7V06VLt3btXvXv3VpcuXXTw4EF3n7Nnz2rGjBlatGiRPv30U7lcLvXr10+S1LdvX40ZM0Z169ZVSkqKUlJS1Ldv3xzvOz09XS6Xy2MDAADFU75+z1T9+vU1efJkSVKtWrU0d+5cbdy4UR07dlSFChUkSWXKlFFwcLD7nBkzZmj8+PEaMmSIJKl69eqaPn26xo4d6x5Lku655x7dd9997v3+/furX79+mjp1qrutQYMGkqRDhw5pyZIl+v7773XddddJkh5//HGtXbtWsbGxmjlzpqQ/HhfOnTtXLVq0kPRHGIyIiNBnn32m5s2bKyAgQD4+Ph715iQqKsqjDgAAUHzle5j6s5CQEP3000+XPOeLL77Qrl27PFaiMjMzde7cOZ09e1b+/v6S5LHqJEnx8fEaNmxYjmPu3r1bxhjVrl3boz09PV3lypVz7/v4+HiMGx4erjJlyigxMVHNmze/ZN1/NmHCBD322GPufZfLpdDQ0FyfDwAAio58DVMlSpTw2LcsS1lZWZc8JysrS1OnTvV43+kCPz8/9z+XKlXK41jJkiUvOaa3t7e++OILeXt7exwLCAjIVuNf5dR2KQ6HQw6H47LOAQAARVOB/jmZEiVKKDMz06OtcePGOnDggGrWrHlZY9WvX18bN27Uvffem+1Yo0aNlJmZqZ9++km33HLLRcfIyMjQ559/7l6FOnDggE6dOqXw8HBJkq+vb7Z6AQDAta1Aw1RYWJg2btyom266SQ6HQ0FBQZo0aZK6d++u0NBQ9e7dW15eXtq7d6/27dvn/q29nEyePFnt27dXjRo11K9fP2VkZGjNmjUaO3asateurQEDBmjw4MGaM2eOGjVqpF9++UWbNm3SjTfeqK5du0r6I9yNHDlSL730kkqUKKGHH35YLVu2dIersLAwHT58WPHx8br++utVunRpVqAAALjGFej3TM2ZM0fr169XaGioGjVqJEnq3LmzPvzwQ61fv17NmjVTy5Yt9dxzz6lq1aqXHKtNmzZ65513tHLlSjVs2FDt2rXTzp073cdjY2M1ePBgjRkzRnXq1NE//vEP7dy50+NdJn9/f40bN0733HOPWrVqpZIlS2rp0qXu47169VKXLl3Utm1bVahQQUuWLMnjGQEAAEWNZYwxBV1EYRAXF6fRo0fr1KlTeT62y+WS0+lUWlqaAgMD83x8oKgJG78qT8ZJju6WJ+MAQE5y+/O7WHwDOgAAQEEp0HemAFybWFECUJywMvX/RUZG5ssjPgAAULwRpgAAAGwgTAEAANhAmAIAALCBMAUAAGADYQoAAMAGwhQAAIANfM/UVXDhS+ZdLlcBVwIAAHLrws/tv/tjMYSpq+D06dOS5PF3AAEAQNFw+vRpOZ3Oix7nb/NdBVlZWTp27JhKly4ty7Ly/Xoul0uhoaE6evQofwswDzCfeYv5zFvMZ95hLvNWcZhPY4xOnz6t6667Tl5eF38zipWpq8DLy0vXX3/9Vb9uYGBgkf0AF0bMZ95iPvMW85l3mMu8VdTn81IrUhfwAjoAAIANhCkAAAAbCFPFkMPh0OTJk+VwOAq6lGKB+cxbzGfeYj7zDnOZt66l+eQFdAAAABtYmQIAALCBMAUAAGADYQoAAMAGwhQAAIANhKliYsaMGWrdurX8/f1VpkyZXJ1jjNGUKVN03XXXqWTJkmrTpo2+/vrr/C20iDh58qQGDRokp9Mpp9OpQYMG6dSpU5c8JzIyUpZleWwtW7a8OgUXMvPmzVO1atXk5+enJk2aaOvWrZfs//HHH6tJkyby8/NT9erVtWDBgqtUadFwOfO5ZcuWbJ9Dy7K0f//+q1hx4fTJJ5+oR48euu6662RZlt57772/PYfP5sVd7nwW588mYaqYOH/+vHr37q1//vOfuT5n1qxZeu655zR37lzt2rVLwcHB6tixo/tvCV7L7rnnHsXHx2vt2rVau3at4uPjNWjQoL89r0uXLkpJSXFvq1evvgrVFi5vv/22Ro8erSeffFJ79uzRLbfcottvv11HjhzJsf/hw4fVtWtX3XLLLdqzZ48mTpyoUaNGafny5Ve58sLpcufzggMHDnh8FmvVqnWVKi68zpw5owYNGmju3Lm56s9n89Iudz4vKJafTYNiJTY21jidzr/tl5WVZYKDg010dLS77dy5c8bpdJoFCxbkY4WFX0JCgpFkduzY4W7bvn27kWT2799/0fOGDBlievbseRUqLNyaN29uHnzwQY+28PBwM378+Bz7jx071oSHh3u0PfDAA6Zly5b5VmNRcrnzuXnzZiPJnDx58ipUV3RJMu++++4l+/DZzL3czGdx/myyMnWNOnz4sFJTU9WpUyd3m8Ph0G233aZt27YVYGUFb/v27XI6nWrRooW7rWXLlnI6nX87N1u2bFHFihVVu3ZtDRs2TD/99FN+l1uonD9/Xl988YXH50qSOnXqdNG52759e7b+nTt31ueff67ff/8932otCq5kPi9o1KiRQkJC1L59e23evDk/yyy2+Gzmj+L42SRMXaNSU1MlSZUqVfJor1SpkvvYtSo1NVUVK1bM1l6xYsVLzs3tt9+ut956S5s2bdKcOXO0a9cutWvXTunp6flZbqHyyy+/KDMz87I+V6mpqTn2z8jI0C+//JJvtRYFVzKfISEheu2117R8+XKtWLFCderUUfv27fXJJ59cjZKLFT6beas4fzZ9CroAXNyUKVM0derUS/bZtWuXmjZtesXXsCzLY98Yk62tuMjtfErZ50X6+7np27ev+5/r1aunpk2bqmrVqlq1apXuuuuuK6y6aLrcz1VO/XNqv1ZdznzWqVNHderUce+3atVKR48e1ezZs3Xrrbfma53FEZ/NvFOcP5uEqULs4YcfVr9+/S7ZJyws7IrGDg4OlvTH/3mFhIS423/66ads/ydWXOR2Pvfu3asff/wx27Gff/75suYmJCREVatW1cGDBy+71qKqfPny8vb2zrZqcqnPVXBwcI79fXx8VK5cuXyrtSi4kvnMScuWLfXmm2/mdXnFHp/N/FdcPpuEqUKsfPnyKl++fL6MXa1aNQUHB2v9+vVq1KiRpD/ez/j444/1zDPP5Ms1C1pu57NVq1ZKS0vTZ599pubNm0uSdu7cqbS0NLVu3TrX1zt+/LiOHj3qEVaLO19fXzVp0kTr16/XnXfe6W5fv369evbsmeM5rVq10gcffODRtm7dOjVt2lQlSpTI13oLuyuZz5zs2bPnmvoc5hU+m/mv2Hw2C/Ltd+Sd7777zuzZs8dMnTrVBAQEmD179pg9e/aY06dPu/vUqVPHrFixwr0fHR1tnE6nWbFihdm3b5/p37+/CQkJMS6XqyBuoVDp0qWLqV+/vtm+fbvZvn27ufHGG0337t09+vx5Pk+fPm3GjBljtm3bZg4fPmw2b95sWrVqZSpXrnzNzefSpUtNiRIlzMKFC01CQoIZPXq0KVWqlElOTjbGGDN+/HgzaNAgd/9vv/3W+Pv7m0cffdQkJCSYhQsXmhIlSphly5YV1C0UKpc7n88//7x59913TVJSkvnqq6/M+PHjjSSzfPnygrqFQuP06dPu/zZKMs8995zZs2eP+e6774wxfDYv1+XOZ3H+bBKmiokhQ4YYSdm2zZs3u/tIMrGxse79rKwsM3nyZBMcHGwcDoe59dZbzb59+65+8YXQ8ePHzYABA0zp0qVN6dKlzYABA7L9Ou+f5/Ps2bOmU6dOpkKFCqZEiRKmSpUqZsiQIebIkSNXv/hC4JVXXjFVq1Y1vr6+pnHjxubjjz92HxsyZIi57bbbPPpv2bLFNGrUyPj6+pqwsDAzf/78q1xx4XY58/nMM8+YGjVqGD8/PxMUFGRuvvlms2rVqgKouvC58Kv5f92GDBlijOGzebkudz6L82fTMub/v00HAACAy8ZXIwAAANhAmAIAALCBMAUAAGADYQoAAMAGwhQAAIANhCkAAAAbCFMAAAA2EKYAAABsIEwBKPYsy9J7771X0GXYdvbsWfXq1UuBgYGyLEunTp3KsS0sLEwvvPBCrsaMi4tTmTJl8rVuoLgjTAGQJEVGRsqyLFmWJR8fH1WpUkX//Oc/dfLkyYIuzbaUlBTdfvvtBV2GbYsWLdLWrVu1bds2paSkyOl05ti2a9cuDR8+PFdj9u3bV0lJSXla55YtW9zBDrgW+BR0AQAKjy5duig2NlYZGRlKSEjQfffdp1OnTmnJkiX5dk1jjDIzM+Xjk3//OQoODs63sa+mQ4cOKSIiQvXq1btkW4UKFXI9ZsmSJVWyZMk8rRO41rAyBcDN4XAoODhY119/vTp16qS+fftq3bp1Hn1iY2MVEREhPz8/hYeHa968eR7Ht23bpoYNG8rPz09NmzbVe++9J8uyFB8fL+l/qxYfffSRmjZtKofDoa1bt8oYo1mzZql69eoqWbKkGjRooGXLlrnHPXnypAYMGKAKFSqoZMmSqlWrlmJjYyVJ58+f18MPP6yQkBD5+fkpLCxMUVFR7nP/+phv3759ateunUqWLKly5cpp+PDh+vXXX93HIyMjdccdd2j27NkKCQlRuXLlNGLECP3++++XnL+VK1eqadOm8vPzU/ny5XXXXXd51D948GAFBQXJ399ft99+uw4ePJht7m699VaVLFlSoaGhGjVqlM6cOSNJatOmjebMmaNPPvlElmWpTZs2ObZJyvaY79SpUxo+fLgqVaokPz8/1atXTx9++KGknB/zffDBB2rSpIn8/PxUvXp1TZ06VRkZGR7z+e9//1t33nmn/P39VatWLa1cuVKSlJycrLZt20qSgoKCZFmWIiMjJUnLli3TjTfe6J73Dh06uO8PKNIK9u8sAygshgwZYnr27OneP3TokLnhhhtMpUqV3G2vvfaaCQkJMcuXLzfffvutWb58uSlbtqyJi4szxhjjcrlM2bJlzcCBA83XX39tVq9ebWrXrm0kmT179hhj/veX5uvXr2/WrVtnvvnmG/PLL7+YiRMnmvDwcLN27Vpz6NAhExsbaxwOh9myZYsxxpgRI0aYhg0bml27dpnDhw+b9evXm5UrVxpjjHn22WdNaGio+eSTT0xycrLZunWrWbx4sbtuSebdd981xhhz5swZc91115m77rrL7Nu3z2zcuNFUq1bN/ZfuL8xFYGCgefDBB01iYqL54IMPjL+/v3nttdcuOn8ffvih8fb2NpMmTTIJCQkmPj7ezJgxw338H//4h4mIiDCffPKJiY+PN507dzY1a9Y058+fN8YYs3fvXhMQEGCef/55k5SUZD799FPTqFEjExkZaYwx5vjx42bYsGGmVatWJiUlxRw/fjzHNmOMqVq1qnn++eeNMcZkZmaali1bmrp165p169aZQ4cOmQ8++MCsXr3aGGNMbGyscTqd7jrXrl1rAgMDTVxcnDl06JBZt26dCQsLM1OmTPGYz+uvv94sXrzYHDx40IwaNcoEBASY48ePm4yMDLN8+XIjyRw4cMCkpKSYU6dOmWPHjhkfHx/z3HPPmcOHD5u9e/eaV155xZw+ffqicwoUFYQpAMaYPwKEt7e3KVWqlPHz8zOSjCTz3HPPufuEhoZ6hBRjjJk+fbpp1aqVMcaY+fPnm3LlypnffvvNffz111/PMUy999577j6//vqr8fPzM9u2bfMYe+jQoaZ///7GGGN69Ohh7r333hxrHzlypGnXrp3JysrK8fifw9Rrr71mgoKCzK+//uo+vmrVKuPl5WVSU1Pdc1G1alWTkZHh7tO7d2/Tt2/fHMc3xphWrVqZAQMG5HgsKSnJSDKffvqpu+2XX34xJUuWNP/973+NMcYMGjTIDB8+3OO8rVu3Gi8vL/d8PvLII+a2227z6JNT25/D1EcffWS8vLzMgQMHcqztr2HqlltuMTNnzvTo85///MeEhIS49yWZf/3rX+79X3/91ViWZdasWWOM+d+/45MnT7r7fPHFF0aSSU5OzrEOoCjjnSkAbm3bttX8+fN19uxZ/fvf/1ZSUpJGjhwpSfr555919OhRDR06VMOGDXOfk5GRIafTKUk6cOCA6tevLz8/P/fx5s2b53itpk2buv85ISFB586dU8eOHT36nD9/Xo0aNZIk/fOf/1SvXr20e/duderUSXfccYdat24t6Y/Hch07dlSdOnXUpUsXde/eXZ06dcrxuomJiWrQoIFKlSrlbrvpppuUlZWlAwcOqFKlSpKkunXrytvb290nJCRE+/btu+jcxcfHe8zLX6/p4+OjFi1auNvKlSunOnXqKDExUZL0xRdf6JtvvtFbb73l7mOMUVZWlg4fPqyIiIiLXvtS4uPjdf3116t27dq56v/FF19o165dmjFjhrstMzNT586d09mzZ+Xv7y9Jql+/vvt4qVKlVLp0af30008XHbdBgwZq3769brzxRnXu3FmdOnXS3XffraCgoCu6L6AwIUwBcCtVqpRq1qwpSXrppZfUtm1bTZ06VdOnT1dWVpYk6fXXX/cIBZLcocMYI8uyPI4ZYy56rQsujL1q1SpVrlzZo5/D4ZAk3X777fruu++0atUqbdiwQe3bt9eIESM0e/ZsNW7cWIcPH9aaNWu0YcMG9enTRx06dPB45+rP9fy1xgv+3F6iRIlsxy7UmZNLvcR9sTn4cy1ZWVl64IEHNGrUqGz9qlSpctGx/87lvlyelZWlqVOnerzvdcGfQ/Llzo+3t7fWr1+vbdu2ad26dXr55Zf15JNPaufOnapWrdpl1QgUNryADuCiJk+erNmzZ+vYsWOqVKmSKleurG+//VY1a9b02C78MAwPD9fevXuVnp7uHuPzzz//2+vccMMNcjgcOnLkSLaxQ0ND3f0qVKigyMhIvfnmm3rhhRf02muvuY8FBgaqb9++ev311/X2229r+fLlOnHiRI7Xio+P93jx+dNPP5WXl1euV29yUr9+fW3cuPGi95eRkaGdO3e6244fP66kpCT3ilPjxo319ddfZ7v/mjVrytfX11Zd33//fa6//qBx48Y6cOBAjnV4eeXuR8aFejMzMz3aLcvSTTfdpKlTp2rPnj3y9fXVu+++e3k3BBRCrEwBuKg2bdqobt26mjlzpubOnaspU6Zo1KhRCgwM1O2336709HR9/vnnOnnypB577DHdc889evLJJzV8+HCNHz9eR44c0ezZsyXpoqtBklS6dGk9/vjjevTRR5WVlaWbb75ZLpdL27ZtU0BAgIYMGaJJkyapSZMmqlu3rtLT0/Xhhx+6g8jzzz+vkJAQNWzYUF5eXnrnnXcUHByc45dRDhgwQJMnT9aQIUM0ZcoU/fzzzxo5cqQGDRrkfsR3JSZPnqz27durRo0a6tevnzIyMrRmzRqNHTtWtWrVUs+ePTVs2DC9+uqrKl26tMaPH6/KlSurZ8+ekqRx48apZcuWGjFihIYNG6ZSpUopMTFR69ev18svv3zFdd1222269dZb1atXLz333HOqWbOm9u/fL8uy1KVLl2z9J02apO7duys0NFS9e/eWl5eX9u7dq3379unpp5/O1TWrVq0qy7L04YcfqmvXripZsqS+/vprbdy4UZ06dVLFihW1c+dO/fzzz1f8+BIoTFiZAnBJjz32mF5//XUdPXpU999/v/79738rLi5ON954o2677TbFxcW5V6YCAwP1wQcfKD4+Xg0bNtSTTz6pSZMmSfJ8RJST6dOna9KkSYqKilJERIQ6d+6sDz74wD22r6+vJkyYoPr16+vWW2+Vt7e3li5dKkkKCAjQM888o6ZNm6pZs2ZKTk7W6tWrc1xJ8ff310cffaQTJ06oWbNmuvvuu9W+fXvNnTvX1jy1adNG77zzjlauXKmGDRuqXbt2HitRsbGxatKkibp3765WrVrJGKPVq1e7H5fVr19fH3/8sQ4ePKhbbrlFjRo10lNPPaWQkBBbdUnS8uXL1axZM/Xv31833HCDxo4dm23V6ILOnTvrww8/1Pr169WsWTO1bNlSzz33nKpWrZrr61WuXFlTp07V+PHjValSJT388MMKDAzUJ598oq5du6p27dr617/+pTlz5hSLL1MFLHOxh/kAkAfeeust3XvvvUpLS+PLIQEUSzzmA5Cn3njjDVWvXl2VK1fWl19+qXHjxqlPnz4EKQDFFmEKQJ5KTU3VpEmTlJqaqpCQEPXu3dvj1+wBoLjhMR8AAIANvIAOAABgA2EKAADABsIUAACADYQpAAAAGwhTAAAANhCmAAAAbCBMAQAA2ECYAgAAsOH/ASLH9kwGsf2rAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Linear regression as surrogate model\n",
    "import pandas as pd\n",
    "import xgboost as xgb\n",
    "from sklearn.linear_model import LinearRegression\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "\n",
    "\n",
    "\n",
    "# Produce (probability) forecasts for the test data\n",
    "yhat_prob = model.predict_proba(X_test)[:, 1]\n",
    "\n",
    "# Fit a linear regression model on the original features using the probability forecasts\n",
    "lreg = LinearRegression()\n",
    "lreg.fit(X_test, yhat_prob)\n",
    "\n",
    "# Create a DataFrame to hold the intercept and coefficients\n",
    "explanation = pd.concat([\n",
    "    pd.DataFrame([lreg.intercept_], index=[\"Intercept\"]),\n",
    "    pd.DataFrame(lreg.coef_, index=X_test.columns)\n",
    "], axis=0)\n",
    "\n",
    "# Plot the regression coefficients\n",
    "explanation.plot.barh(legend=False)\n",
    "plt.xlabel(\"Regression coefficients\")\n",
    "plt.savefig('surrogate_model.png', transparent=True, bbox_inches='tight', dpi=300)\n",
    "plt.show()\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Interpretation of the Surrogate Model Results:\n",
    "\n",
    "The surrogate model plot you provided shows the coefficients of a linear regression model fitted to the probability forecasts from an XGBoost classifier. Each bar represents the magnitude and direction of the impact that each feature has on the predicted probability. Positive coefficients indicate a positive influence on the probability, while negative coefficients indicate a negative influence.\n",
    "\n",
    "Let's interpret the coefficients:\n",
    "\n",
    "1. **DEBTINC**: This feature has a large positive coefficient, suggesting that as the debt-to-income ratio increases, the probability of the target variable (`BAD`) being positive increases.\n",
    "\n",
    "2. **CLNO**: The number of credit lines has a moderate positive impact on the probability.\n",
    "\n",
    "3. **DEROG**: The number of derogatory reports has a moderate positive impact on the probability, indicating that more derogatory marks increase the likelihood of the target variable being positive.\n",
    "\n",
    "4. **CLAGE**: The age of the oldest credit line has a large negative impact on the probability, indicating that older credit lines are associated with a lower probability of the target variable being positive.\n",
    "\n",
    "5. **DELINQ**: The number of delinquent credit lines has a large positive impact on the probability, suggesting that more delinquencies significantly increase the likelihood of the target variable being positive.\n",
    "\n",
    "6. **YOJ**: Years at the current job has a very small positive impact on the probability.\n",
    "\n",
    "7. **VALUE**: The value of the property has a small positive impact on the probability.\n",
    "\n",
    "8. **MORTDUE**: The amount due on the mortgage has a small negative impact on the probability.\n",
    "\n",
    "9. **LOAN**: The loan amount has a small negative impact on the probability.\n",
    "\n",
    "10. **NINQ**: The number of recent credit inquiries has a small negative impact on the probability.\n",
    "\n",
    "11. **Intercept**: The intercept of the linear regression model, representing the baseline probability when all feature values are zero, is positive.\n",
    "\n",
    "### Key Takeaways:\n",
    "- **DEBTINC** and **DELINQ** are the most influential features with large positive impacts, indicating that higher debt-to-income ratios and more delinquencies significantly increase the probability of the target variable being positive.\n",
    "- **CLAGE** has a significant negative impact, suggesting that older credit lines reduce the probability of the target variable being positive.\n",
    "- Other features like **CLNO** and **DEROG** also contribute positively but to a lesser extent.\n",
    "- Features such as **MORTDUE**, **LOAN**, and **NINQ** have smaller negative impacts, indicating that higher values in these features slightly decrease the probability of the target variable being positive.\n",
    "\n",
    "This interpretation helps understand which features most strongly influence the model's predictions, offering insights into the relationship between the input features and the predicted probabilities."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "oSMD5hnp3zfV"
   },
   "source": [
    "Given that the linear regression can give no more than an approximation of the neural network, it is crucial to examine the quality of the approximation. If it is poor, the insights, which we expect the surrogate model to provide may not be reliable. \n",
    "\n",
    "Write some code to assess the quality of the approximation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 216,
     "status": "ok",
     "timestamp": 1680274419193,
     "user": {
      "displayName": "Stefan Lessmann",
      "userId": "06342662613942148717"
     },
     "user_tz": -120
    },
    "id": "o7CC2AiC155C",
    "outputId": "9ebf5e7a-06c0-4e37-8a0a-09f24741104d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "XGB probability forecast:  [0.01944919]\n",
      "LR approximation:  [0.15325803]\n"
     ]
    }
   ],
   "source": [
    "# Compare XGB predictions to linear regression approximation\n",
    "\n",
    "\n",
    "test_point = X_test.iloc[0:1]  # some test point\n",
    "\n",
    "print(\"XGB probability forecast: \", model.predict_proba(test_point)[:, 1])\n",
    "print(\"LR approximation: \", lreg.predict(test_point))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "DGTazxWTwojZ"
   },
   "source": [
    "To examine the agreement between our XGB and regression-based surrogate we can also compare their predictions for the entire test set, and plot the result for better readability."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 423
    },
    "executionInfo": {
     "elapsed": 242,
     "status": "ok",
     "timestamp": 1680274425496,
     "user": {
      "displayName": "Stefan Lessmann",
      "userId": "06342662613942148717"
     },
     "user_tz": -120
    },
    "id": "b-CMrzvo7b0p",
    "outputId": "7187f7b3-6224-4a85-8a08-1eec0dadcec7"
   },
   "outputs": [
    {
     "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>Y</th>\n",
       "      <th>XGB</th>\n",
       "      <th>LR</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1344</th>\n",
       "      <td>0</td>\n",
       "      <td>0.019449</td>\n",
       "      <td>0.153258</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>625</th>\n",
       "      <td>0</td>\n",
       "      <td>0.003941</td>\n",
       "      <td>0.083905</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5908</th>\n",
       "      <td>0</td>\n",
       "      <td>0.201364</td>\n",
       "      <td>0.249744</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2991</th>\n",
       "      <td>0</td>\n",
       "      <td>0.001328</td>\n",
       "      <td>-0.028133</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1545</th>\n",
       "      <td>0</td>\n",
       "      <td>0.000870</td>\n",
       "      <td>0.095415</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3688</th>\n",
       "      <td>0</td>\n",
       "      <td>0.054867</td>\n",
       "      <td>0.207586</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2404</th>\n",
       "      <td>0</td>\n",
       "      <td>0.003324</td>\n",
       "      <td>0.216264</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1173</th>\n",
       "      <td>0</td>\n",
       "      <td>0.270792</td>\n",
       "      <td>0.204980</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3209</th>\n",
       "      <td>0</td>\n",
       "      <td>0.000765</td>\n",
       "      <td>0.126292</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2728</th>\n",
       "      <td>0</td>\n",
       "      <td>0.015110</td>\n",
       "      <td>0.242236</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1192 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      Y       XGB        LR\n",
       "1344  0  0.019449  0.153258\n",
       "625   0  0.003941  0.083905\n",
       "5908  0  0.201364  0.249744\n",
       "2991  0  0.001328 -0.028133\n",
       "1545  0  0.000870  0.095415\n",
       "...  ..       ...       ...\n",
       "3688  0  0.054867  0.207586\n",
       "2404  0  0.003324  0.216264\n",
       "1173  0  0.270792  0.204980\n",
       "3209  0  0.000765  0.126292\n",
       "2728  0  0.015110  0.242236\n",
       "\n",
       "[1192 rows x 3 columns]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# DataFrame with true outcomes, nn predictions, and surrogate model approximations\n",
    "df_comp = pd.DataFrame(\n",
    "    {\n",
    "      \"Y\": y_test,  \n",
    "     \"XGB\": yhat_prob,\n",
    "     \"LR\": lreg.predict(X_test)   \n",
    "    }\n",
    ") \n",
    "df_comp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 449
    },
    "executionInfo": {
     "elapsed": 661,
     "status": "ok",
     "timestamp": 1680274800144,
     "user": {
      "displayName": "Stefan Lessmann",
      "userId": "06342662613942148717"
     },
     "user_tz": -120
    },
    "id": "0KsrlGa65m53",
    "outputId": "b51dd932-a19c-4568-e214-f0d806e4a1e9"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Scatter plot of regression versus XGB outputs, colored by actual target\n",
    "# Plot comparison\n",
    "plt.scatter(df_comp.LR[df_comp.Y == 0], df_comp.XGB[df_comp.Y == 0], label=\"Goods\")\n",
    "plt.scatter(df_comp.LR[df_comp.Y == 1], df_comp.XGB[df_comp.Y == 1], label=\"Bads\")\n",
    "plt.xlabel(\"Regression approximation\")\n",
    "plt.ylabel(\"XGB probability forecast\")\n",
    "plt.legend(loc=\"best\")\n",
    "plt.savefig('reg_XGB.png', transparent=True, bbox_inches='tight', dpi=300)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Interpretation of the Plot:\n",
    "\n",
    "The scatter plot shows the relationship between the regression approximation (from the linear regression surrogate model) and the XGBoost probability forecast for two classes: `Goods` (blue dots) and `Bads` (orange dots).\n",
    "\n",
    "#### Key Observations:\n",
    "1. **XGBoost Probability Forecast**: The vertical axis represents the probability predictions from the XGBoost model. These probabilities range from 0 to 1, where higher values indicate a higher likelihood of being classified as `BAD`.\n",
    "\n",
    "2. **Regression Approximation**: The horizontal axis represents the predictions from the linear regression model, which approximates the XGBoost probability forecasts.\n",
    "\n",
    "3. **Distribution of Goods (blue dots)**:\n",
    "    - Most `Goods` are clustered around a lower probability forecast (near 0) from the XGBoost model.\n",
    "    - The linear regression approximations for `Goods` also tend to be low (around 0).\n",
    "\n",
    "4. **Distribution of Bads (orange dots)**:\n",
    "    - Many `Bads` are clustered at a high probability forecast (near 1) from the XGBoost model.\n",
    "    - The linear regression approximations for `Bads` are spread across a wider range but still show a tendency to be higher than those for `Goods`.\n",
    "\n",
    "5. **Alignment of Points**:\n",
    "    - There is a noticeable spread in the `Regression approximation` axis, suggesting that while the linear regression model captures some of the trends in the XGBoost predictions, there is still some discrepancy.\n",
    "\n",
    "6. **Overlap of Goods and Bads**:\n",
    "    - There is some overlap between `Goods` and `Bads` in the lower probability forecast range (0 to 0.4), indicating that both classes can share similar probability forecasts and approximations in this range.\n",
    "    - `Bads` are more spread out in higher probability forecasts, indicating more variability in the linear regression model's approximation.\n",
    "\n",
    "### Conclusion:\n",
    "- The linear regression surrogate model captures the general trend of the XGBoost probability forecasts but with some variability.\n",
    "- The XGBoost model seems effective in distinguishing between `Goods` and `Bads`, especially at the higher probability forecasts.\n",
    "- The surrogate linear regression model provides an approximation that can help in understanding the feature contributions but may not be as precise as the XGBoost model in capturing the exact probability forecasts. \n",
    "\n",
    "This plot helps visualize how well the linear regression model approximates the XGBoost model's predictions and highlights the areas where the approximation might be less accurate."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "kZm4SJeyIgw3"
   },
   "source": [
    "## Permutation-based feature importance\n",
    "We introduces the concepts of permutation-based feature importance in the lecture. Now we want to examine how it works in Python. Of course, this will require us finding a Python method that implements the approach. To shortcut your searching, note that there is such method in *the most popular Python library for ML* and it is called `permutation_importance`. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "executionInfo": {
     "elapsed": 578,
     "status": "ok",
     "timestamp": 1680275036502,
     "user": {
      "displayName": "Stefan Lessmann",
      "userId": "06342662613942148717"
     },
     "user_tz": -120
    },
    "id": "S8syUjAMK-i5"
   },
   "outputs": [],
   "source": [
    "# Permutation-based feature importance\n",
    "from sklearn.inspection import permutation_importance\n",
    "feature_imp = permutation_importance(estimator=model, X=X_test, y=y_test)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "1gZA7O03LNVu"
   },
   "source": [
    "The calculation of feature importance is pretty easy. However, visualizing the estimates importance scores takes a little work. Fortunately, this is a standard task and ready-to-use demo code is easily found on the web. In fact, we only found to browse the [official sklearn website](https://scikit-learn.org/stable/auto_examples/inspection/plot_permutation_importance_multicollinear.html#sphx-glr-auto-examples-inspection-plot-permutation-importance-multicollinear-py) and will find some nice demo code to plot feature importance scores as a boxplot.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 430
    },
    "executionInfo": {
     "elapsed": 544,
     "status": "ok",
     "timestamp": 1680275047105,
     "user": {
      "displayName": "Stefan Lessmann",
      "userId": "06342662613942148717"
     },
     "user_tz": -120
    },
    "id": "YvW75YwTLBdJ",
    "outputId": "9079e4a9-5d9a-4fa0-9dca-af10dee244bf"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Creating PDPs \n",
    "perm_sorted_idx = feature_imp.importances_mean.argsort()\n",
    "\n",
    "plt.boxplot(\n",
    "    feature_imp.importances[perm_sorted_idx].T, vert=False, labels=X.columns[perm_sorted_idx]\n",
    ")\n",
    "plt.savefig('feature_importance.png', transparent=True, bbox_inches='tight', dpi=300)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Interpretation of Permutation-Based Feature Importance Chart\n",
    "\n",
    "The box plot shows the distribution of feature importance values for a set of features (DEBTINC, DELINQ, CLNO, CLAGE, VALUE, DEROG, MORTDUE, LOAN, YOJ, NINQ) across multiple permutations of the data. Each box represents the variability in importance scores, providing insights into which features are consistently important and which ones have more variable importance.\n",
    "\n",
    "#### Key Elements of the Box Plot:\n",
    "1. **Boxes**: Each box represents the interquartile range (IQR) of the feature importance values, i.e., the middle 50% of the data. The bottom and top edges of the box represent the 25th and 75th percentiles, respectively.\n",
    "2. **Median Line**: The orange line inside each box represents the median (50th percentile) of the feature importance values.\n",
    "3. **Whiskers**: The lines extending from the boxes (whiskers) represent the range of the data, excluding outliers. Typically, the whiskers extend to 1.5 times the IQR from the quartiles.\n",
    "4. **Outliers**: Individual points beyond the whiskers represent outliers in the data.\n",
    "\n",
    "### Interpretation of Each Feature:\n",
    "- **DEBTINC**: The box plot shows a relatively narrow IQR with a high median, indicating that DEBTINC is consistently an important feature with one notable outlier.\n",
    "- **DELINQ**: This feature has a relatively narrow IQR with a high median, indicating consistent importance with a few outliers.\n",
    "- **CLNO**: The number of credit lines shows a moderate IQR, indicating variability in its importance across different permutations. The whiskers and a few outliers suggest a broader range of values.\n",
    "- **CLAGE**: The age of the oldest credit line shows a moderate IQR, suggesting variability in its importance. Outliers are present.\n",
    "- **VALUE**: The property value shows a moderate IQR with a few outliers, indicating variability in its importance.\n",
    "- **DEROG**: This feature has a narrow IQR, indicating consistent importance with minimal outliers.\n",
    "- **MORTDUE**: The amount due on the mortgage has a relatively narrow IQR, indicating consistent importance with minimal outliers.\n",
    "- **LOAN**: The loan amount shows a narrow IQR, indicating consistent importance among most observations.\n",
    "- **YOJ**: Years at the current job has a very narrow IQR, suggesting that its importance is consistent with few outliers.\n",
    "- **NINQ**: The number of recent credit inquiries shows a narrow IQR, indicating consistent importance with some variability and outliers.\n",
    "\n",
    "### Conclusion:\n",
    "The box plot provides a clear visualization of the distribution and spread of each feature's importance in the dataset. Features like DEBTINC and DELINQ have consistently high importance with some variability, while features like LOAN and YOJ are more consistent with fewer outliers. Understanding these distributions helps in interpreting the importance and impact of each feature in predictive modeling.\n",
    "\n",
    "### Key Takeaways:\n",
    "- **Consistently Important**: DEBTINC, DELINQ, and CLNO show high median importance values, indicating they are crucial in the model's predictions.\n",
    "- **Variable Importance**: Features like CLAGE and VALUE have more variability, suggesting their importance can change depending on the data permutation.\n",
    "- **Less Important**: Features like YOJ, LOAN, and NINQ have lower median importance values, indicating they have less influence on the model's predictions.\n",
    "\n",
    "This interpretation helps in understanding which features are consistently important and which ones have more variable importance, aiding in model interpretability and feature selection."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "6XKus66XIhEL"
   },
   "source": [
    "## Partial dependence\n",
    "Next up, we discussed the concept of partial dependence. A partial dependence plot (PDP) complements feature importance analysis. For example, it is interesting to examine how those features that importance analysis identified as crucial affect model predictions (when averaging over the effect of the other features). Let's try to create a PDP for our XGB-classifier. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 449
    },
    "executionInfo": {
     "elapsed": 2145,
     "status": "ok",
     "timestamp": 1680175374531,
     "user": {
      "displayName": "Stefan Lessmann",
      "userId": "06342662613942148717"
     },
     "user_tz": -120
    },
    "id": "D7f9vg2_IhOp",
    "outputId": "440b2dfe-cd9d-4988-8862-75cd628622a2"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Partial dependence analysis\n",
    "from sklearn.inspection import PartialDependenceDisplay\n",
    "\n",
    "PartialDependenceDisplay.from_estimator(estimator=model,X=X_test, features=['MORTDUE'])\n",
    "plt.savefig('partial_dependence.png', transparent=True, bbox_inches='tight', dpi=300)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ajz6Wd1FT21b"
   },
   "source": [
    "### Interpretation of the Partial Dependence Plot:\n",
    "\n",
    "The partial dependence plot (PDP) shows the marginal effect of the `MORTDUE` feature (the amount due on the mortgage) on the predicted outcome. This type of plot helps us understand how the model's predictions change when the `MORTDUE` feature varies, while all other features are held constant.\n",
    "\n",
    "#### Key Elements of the PDP:\n",
    "1. **X-Axis (MORTDUE)**: This represents the values of the `MORTDUE` feature. It ranges from approximately 0.05 to 0.35.\n",
    "2. **Y-Axis (Partial Dependence)**: This shows the predicted outcome or probability, ranging from approximately 0.18 to 0.26.\n",
    "3. **Ticks on X-Axis**: These indicate the distribution of the `MORTDUE` values in the dataset. The density of ticks gives a sense of how frequently different values of `MORTDUE` occur.\n",
    "\n",
    "### Key Observations:\n",
    "1. **Overall Trend**: The plot shows that the partial dependence of `MORTDUE` on the predicted outcome fluctuates significantly, indicating that `MORTDUE` has a complex relationship with the predicted probability.\n",
    "2. **High Partial Dependence for Low `MORTDUE`**: When `MORTDUE` is low (around 0.05 to 0.10), the partial dependence is relatively high, close to 0.26. This suggests that lower amounts due on the mortgage are associated with higher predicted probabilities.\n",
    "3. **Decrease in Partial Dependence**: As `MORTDUE` increases from around 0.10 to 0.20, the partial dependence drops significantly, reaching a low around 0.18. This indicates that medium values of `MORTDUE` are associated with lower predicted probabilities.\n",
    "4. **Increase and Fluctuations**: Beyond 0.20, the partial dependence increases again but shows fluctuations. At around 0.25 to 0.30, there is an increase in the partial dependence to around 0.24, followed by a slight decrease and then stabilization.\n",
    "5. **High Partial Dependence for High `MORTDUE`**: For higher values of `MORTDUE` (around 0.30 and above), the partial dependence stabilizes at around 0.23 to 0.24.\n",
    "\n",
    "### Conclusion:\n",
    "- **Complex Relationship**: The `MORTDUE` feature has a non-linear and complex relationship with the predicted outcome.\n",
    "- **Impact at Different Ranges**: Low `MORTDUE` values are associated with higher predicted probabilities, while medium `MORTDUE` values have lower predicted probabilities. High `MORTDUE` values again show an increase in predicted probabilities, but with fluctuations.\n",
    "- **Insightful Feature**: Understanding the partial dependence of `MORTDUE` can provide valuable insights into how changes in the amount due on the mortgage influence the model's predictions.\n",
    "\n",
    "This PDP helps in interpreting the significance of the `MORTDUE` feature and its impact on the prediction, offering a deeper understanding of the model's behavior."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}