Final YearAdvancedAI/ML 3-4 months
Disease Prediction from Symptoms
A clinical-decision-support app that predicts likely conditions from reported symptoms and history, with calibrated probabilities and clear disclaimers.
Pythonscikit-learn / XGBoostFastAPIReactSHAP
Find GitHub repos & source code →How to build it — step by step
- 1Data: Use a curated symptom-disease dataset; clean, encode, and handle class imbalance.
- 2Modelling: Train and compare classifiers; calibrate probabilities and validate carefully.
- 3Explainability: Use SHAP to show which symptoms drove each prediction.
- 4App + safety: Serve predictions with confidence and prominent "not medical advice" guidance.
Key features to implement
- ✓Symptom-based prediction
- ✓Calibrated probabilities
- ✓SHAP explanations
- ✓Top-N likely conditions
- ✓Clear safety disclaimers
💡 Unique twist to stand out
Add an active-questioning flow that asks the most informative follow-up symptom to reduce uncertainty, like a triage bot.
🎓 What you'll learn
Supervised learning, probability calibration, model explainability, and responsible ML deployment.