Heart Disease Risk Identification using Machine Learning Techniques for a Highly Imbalanced Dataset: a Comparative Study

Heart disease has become one of the most prevailing universal diseases in world today. It is estimated that 32% all deaths worldwide are caused due to heart diseases. One major causes for this its extremely difficult even medical practitioners predict as attacks it a complex task which requires great amount knowledge and experience. The number by hugely increased recent past. Machine learning popular areas computer science where many problems have been addressed successfully specially field medicine. In study we trained multiple supervised classifiers namely’; Naïve Bayes, LightGBM, Decision Trees, Random Forest, XGBoost, K Nearest Neighbours ADABoost compared accuracies identified what models perform better prediction. We used Behavioral Risk Factor Surveillance System (BRFSS) 2015 Disease Health Indicators Dataset was highly imbalanced order address class imbalance problem methods such Synthetic Minority Over Sampling Technique (Smote) Sampling, Adaptive Under TomekLink, SmoteTomek, Smoteen Cluster Centroid. According results obtained, can conclude hybrid SmoteTomek performed than other sampling methods.