Skip to content

Predicting Presence of Heart Disease

This project looks at whether everyday clinical measurements can be used to help identify patients who may have heart disease. Using a public dataset of just anonymous patient records, it explores how information like chest pain type, blood pressure, heart rate, and stress test results can be combined to support clinical decision-making, with the goal of highlighting higher-risk cases rather than replacing medical judgment.

The project compared a simple, easy-to-interpret Logistic Regression model with a more advanced machine learning approach (Gradient Boosting ensemble) to see how much predictive performance could be improved. The more flexible model performed significantly better, showing that patterns in routine health data can be used to make surprisingly accurate predictions, while also reinforcing the importance of using these tools carefully and ethically in healthcare settings.

Working Code