Predicting the Cardiac Diseases using SelectKBest Method Equipped Light Gradient Boosting Machine
Manikandan Ayyanar, Selvaprabu Jeganathan, Saravanan Parthasarathy, Vaishnavi Jayaraman, L. Arun Raj
Abstract
Cardiovascular diseases cause a significant number of deaths around the world. It is vital to examine the data generated by the healthcare industry to find the hidden pattern. In this study, a cardiovascular dataset was scrutinized by correlating the risk factors to generate different analytical insights which portray the risk population entitled to cardiac diseases. The boosting techniques such as Light Gradient Boosting Machine, AdaBoost, CatBoost, LogitBoost, and XGBoost were applied to predict the likelihood of cardiac diseases. The performances of the boosting techniques were compared based on the Accuracy, Precision, Recall, F1 and AUC score, and Run time. The performance of the boosting classifiers was improved by employing the SelectKBest feature selection method. On the interpretation of the results obtained from the five boosting models before and after the application of the feature selection method, the Light Gradient Boosting Machine (LGBM) model performed well. The LGBM model with ten independent variables predicted the occurrence of cardiac disease with 73.63% accuracy and 0.8016 of AUC value.