A REAL TIME HEART FAILURE PREDICTION TECHNIQUE USING DATA MINING AND SMOTE METHOD

Abstract

The significant global mortality & morbidity are attributable to the cardiovascular-CVD disease. Predicting who will or will not survive cardiac illness is an enormous issue in clinical data analytics. The health business generates vast volumes of raw data, and data mining could take this data and turn it into insights that help guide decision-making. Numerous investigations demonstrated that crucial elements help enhance ML model performance. Its research looks at the 299 hospitalized individuals who survived cardiac failure. The goal is to improve the accuracy of survivor prediction for cardiovascular patients by identifying important variables & effective data mining methods. This research uses nine classification models to predict patient survival: DT, Adaptive boosting classifier (AdaBoost), LR, SGD, RF, GBM, ETC, G-NB, & SVM.  The synthetic Minority Oversampling Method addresses the issue of class imbalance (SMOTE).The characteristics with the highest rankings chosen by RF are then used to train ML models. Machine learning-ML methods that use a more comprehensive set of features are also used for comparison. According to the data obtained from the experiments, ETC is the most effective model for predicting the survival of heart patients, with an accuracy value of 0.9262 compared to SMOTE.