MediCare

Description

💡 MediCare — Disease Prediction System

This project implements a Machine Learning-based Disease Prediction System designed to predict probable diseases based on a patient's symptoms. The core model is an optimized XGBoost Classifier (XGBClassifier), fine-tuned using RandomizedSearchCV for enhanced accuracy and performance.

⚙️ Tech Stack

Language: Python
Libraries:
pandas, numpy, scikit-learn, xgboost, joblib, difflib, tabulate, scipy

🧠 Model Summary

Model Used: XGBClassifier (eXtreme Gradient Boosting)
Training Strategy:
- Features selected using SelectKBest with mutual_info_classif.
- Dataset split into training and testing sets using stratification.
- Hyperparameters optimized via RandomizedSearchCV.
- Final model saved as optimized_disease_predictor.pkl.

🔍 Features

Interactive symptom input system with fuzzy matching (difflib) for typo correction.
Dynamic prediction output showing:
- Primary predicted disease with confidence score.
- Alternative likely diseases.
Symptom data structured for binary classification (1 = present, 0 = absent).
Model trained on selected 40 most informative symptoms for high relevance and generalization.

💡 Prediction Output

Predicts top three possible diseases with confidence percentages.
Offers clear suggestions and highlights the need for professional medical consultation.

Progress During Hackathon

💡 Project Name: MediCare 🩺 Tagline / Slogan: Predict Prevent Protect ✅ Current Progress: Model Development: Symptom-based Disease Prediction system built. Trained an XGBoost Classifier on real-world medical symptom datasets. Feature selection using Mutual Information for improving prediction accuracy. Model Accuracy: Achieved a strong predictive accuracy (~High 90% range depending on dataset split). Functionality: Predicts the top 3 probable diseases based on user-reported symptoms. Includes both interactive input and predefined input support. Intelligent fuzzy symptom matching (helps with typos or approximate inputs). Tech Stack: Python, Pandas, Scikit-learn, XGBoost, Joblib Difflib for close symptom matching. Tabulate for clean terminal output. Deployment Preparation: Model is saved and can be integrated into web / desktop apps. CLI prototype working perfectly for demonstration. ⚡ Next Steps: 🌐 Web / App Integration (suggest: Streamlit / Flask for web app). 🧠 Expand dataset for more diverse symptoms and conditions. 📊 Add more context-aware suggestions (e.g. severity ranking).