Improving Prediction of Chronic Heart Failure using SMOTE and Machine Learning

Around the world, cardiovascular disease (CVD) is a vital reason for death and morbidity.Predicting heart disease survivors is a significant challenge in clinical data analytics.Machine learning converts massive volumes of raw data created by the healthcare business into meaningful knowledge that can aid in decision-making. Several research have shown that important attributes are important in increasing the accuracy of machine learning approaches. This paper looks at heart failure survivors from a group of 299 individuals who were hospitalised to the hospital. The goal is to use of machine learning models that can enhance the predictability of cardiac patient survival. Thispaper employs eight classification models: Random Forest (RM), Extra Tree (ET), NaiveBayes (NB), K Nearest Neighbour (KNN), Decision Tree J48, Decision Table/Naive Bayes hybrid classifier (DTNB), optimized Forest, Alternating Decision Tree (ADTree) to predict patient’s survival. Synthetic Minority Oversampling Technique (SMOTE) is used to resolve the unbalance dataset. Experiment outcomes demonstrate SMOTE technique improves the accuracy of the selected classifier’s output and DTNB achieves highest accuracywith 87.08% utilising SMOTE to forecast the survival of cardiac patients. All experiments are carried out in a simulation environment using the WEKA tool.