← Work
Machine learning
Heart Disease Prediction
Supervised classification model at 89% accuracy, using feature selection and PCA to cut dimensionality, served as a real-time inference app.
- Classification
- Python
- Scikit-learn
Systems implemented
- Supervised classification pipeline reaching 89% accuracy, with preprocessing, feature selection and evaluation treated as one reproducible sequence rather than ad-hoc notebook steps.
- PCA for dimensionality reduction, trading a small amount of variance for a model that trains faster and generalises better than one fed every raw column.
- Feature selection ahead of fitting, so the model is judged on signal rather than on the number of columns it was handed.
- Deployed for real-time inference with Streamlit, tunnelled over Ngrok — the step that turns a notebook into something another person can actually use.
Built in Python with Scikit-learn.