Description

• This is a proposed medical chatbot deployed into DialogFlow that takes user's input, analyzes it and diagnoses the case into a disease from a pre-defined dataset.
• The bot analyses the user's sentence by checking whether it's medically relevant or not, then if it finds the input relevant, it goes to the next step.
• The bot then check for the number of the symptoms found -if any- in the user's inquiry, if they match the threshold then the model is called to send the predicted diagnosis based on these symtoms,
  otherwise the bot starts to ask follow-up questions for the user comprising the most relevant symptom to what he has mentioned before.

Dependencies

numpy==1.18.1.
scikit-learn==0.24.2.
pandas==1.0.1.
nltk==3.4.5.
joblib==0.14.1.
pickle version==4.0.

Files Description

chatbot/disease_classifier.py

Python file that contains the classification code (using SVM only) for the disease which the input symptoms belong to.

chatbot/get_next_symptom.py

Python file that contains a function that takes some symptoms and predicts the next most important symptom to be asked as a follow-up question for the user.

chatbot/medical_relevence.py

Python file that contains a function that takes user's message and decide if it's relevent to the medical context or not.

chatbot/clean_data.csv

Csv file that contains a subset of the clean data used for training the model.

chatbot/disease_classifier.pkl

The SVM classification model used in disease_classifier.py serialized using pickle.

chatbot/medical_relevence_classifier.joblib

The model used in medical_relevence.py

medical_relevence_classification.ipynb

Python notebook for the medical relevance function and its needed steps.

Disease_Prediction_Clustering_NextSymptom.ipynb

Python notebook for different clustering techniques applied (K-means and DBSCAN) and get_next_symptom function applied.

Disease_Prediction_Classification,ipynb

Python notebook for different classification techniques applied (Random forest, XGBoost and SVM)