Muralidharan Jayaraman

Cardiovascular diseases (CVDs) pose a major global health challenge despite medical advancements. Machine Learning (ML) offers potential for early detection and prediction of CVD risk. This research utilized two datasets, Kaggle’s CVD dataset (70,000 instances) and UCI’s CHD dataset (303 instances), for ML model evaluation. After Exploratory Data Analysis (EDA) and data cleansing, the CVD dataset was reduced by 2.67%. Feature engineering introduced new variables like BMI, MAP, and PP to enhance predictive accuracy. The goal is to identify the best ML models. A set of Regression Models in the categories of Linear Regression, Tree Regression, Optimizable Neural Network, Kernel Regression, and Ensemble Regression were deployed on both CVD and CHD datasets. Boosted Tree Ensemble Regression gave an accuracy of 97.37% for CVD dataset, and Robust Linear Regression returned an accuracy of 98.52% for CHD dataset.