AI/MLEarlier work · 2023 – 2026
Heart Disease Risk Prediction
Sixth-semester coursework from 2023 that compared KNN, SVM, Random Forest, a voting ensemble and XGBoost on the widely used 1,025-row heart-disease dataset, later packaged as a Streamlit app that turns the model’s probability into a low / moderate / high risk band.
Technology stack
Pythonscikit-learnXGBoostpandasStreamlit
01
What it is
- A training notebook comparing five classifiers on structured clinical features such as age, chest-pain type, resting blood pressure, cholesterol, maximum heart rate and ST depression.
- A Streamlit app that loads a single joblib artifact, builds the feature vector in a fixed order, and shows the predicted probability as a risk band and gauge.
02
Why no accuracy is shown
The dataset has 1,025 rows but only 302 unique ones, and the notebook splits without de-duplicating, so most test rows have an identical copy in the training split. The accuracies it reports (up to 89.42%) are therefore not a reliable estimate of performance on new patients, and the probabilities are not calibrated.
Sources
- Source code
- Streamlit appsleeps when idle
Checked against the repositories on 22 September 2026.