Explainable Heart Disease Risk Prediction System

Audience

Who it is for

ML majors who need evals they can defend, plus a small web app.

Fill 13 clinical fields, see risk band, top features, and a model card (data, metrics, limits).

Branches

Team size: 1–4 students

Problem

Existing vs proposed

Many student “heart disease” projects stop at 90% accuracy in a notebook. Panels now ask about leakage, class balance, and who is harmed by false negatives.

Full ML pipeline + comparison of logistic regression vs tree ensemble, chosen threshold for recall, SHAP bar for one patient, audit log of predictions.

Build

Modules

1. 1. Data

   • Cleaning, encoding, train/test by patient
   • Imbalance strategy

2. 2. Models

   • Baseline vs ensemble
   • CV, ROC/PR, confusion at two thresholds

3. 3. Explain

   • SHAP global + local
   • Model card

4. 4. App

   • Role login, form, history
   • “Not a diagnosis” banner

Deliverables

Included with the kit

• Complete source code (ZIP + GitHub-style repo)
• Database schema and sample data / dataset notes
• README.md — how to run, stack, modules, and demo path (Windows + college lab)
• Viva question bank with short answers
• Two mentor reviews: architecture, then final demo