A Modern Comparison of ML Algorithms for Cardiovascular Disease Prediction

In today's world, with rise in number of people following a sedentary lifestyle, heart disease has become the top reason of death. It is tough for healthcare personnel to forecast such an ailment in advance because it is a complex undertaking that demands competence and a greater knowledge for prediction. We aimed this study towards looking into the topic of predicting cardiovascular disease in advance with use of different ML models, and comparing the accuracies of different older as well newer techniques, on the Framingham heart study dataset from Kaggle. We have investigated the cardiovascular disease prediction using several techniques, namely, Decision Tree, KNN, Naϊve Bayes, SVM, XGBoost, and Random Forest. The prediction accuracy, ROC curve, and AUC value are all used to evaluate the performance of these machine learning techniques. Based on accuracy scores, KNN, SVM, and RFC perform equally well while scoring an accuracy of 85.33%. However, an analysis of the ROC curve and AUC value provides us with a different picture of how despite slightly lesser accuracy percentages, Naϊve Bayes, the modern method XGBoost, and Random Forest actually perform better than the techniques anticipated earlier based on accuracy alone.