Feature Selection and Prediction of Heart diseases using Gradient Boosting Algorithms

According to WHO, 31% of the global human mortality rate is due to Cardio Vascular Diseases and 85% of it is due to heart attacks and strokes. To prevent such deaths, several Machine Learning (ML) Algorithms are used in the early prediction of heart attacks. In order to reduce the computation time of the ML models, several feature selection techniques exist. In this paper, Feature Importance ranking of two gradient boosting algorithms XGBoost and CatBoost were computed on Cleveland, Statlog heart and SA heart data sets. With each feature importance rank as threshold, subsets of features were formed. Classifiers XGBoost, CatBoost and Majority voting ensemble were modelled on these subsets and the feature subset yielding highest accuracy was obtained. The range of feature importance ranking among which the feature subset with the highest accuracy would be obtained, was identified in this work. The classifiers exhibited improved performance on selected features when compared to their performance on all features. On comparing the classifiers, CatBoost outperformed the other classifiers.