Heart Attack Prediction

Built a clinical risk assessment model predicting heart attack likelihood from 303 patient records with 14 clinical features (age, cholesterol, blood pressure, max heart rate, etc.). Applied RobustScaler for outlier-resistant scaling, IQR-based outlier removal, one-hot encoding, and feature selection via correlation analysis. Compared 4 models with cross-validation. Best result: Logistic Regression at 85.0% accuracy (10-fold CV: 85.1%), tuned with GridSearchCV.

Tech: Python, Scikit-learn, Pandas, Seaborn, Matplotlib