GA-SLE: A hybrid algorithm for heart disease prediction using feature selection mechanism

Abstract

Heart disease is the leading cause of death around the globe, killing more people than cancer. A study published in the Journal of the American Heart Association shows that gene variants in our genes can increase our risk of heart failure by as much as 40%. To make matters worse, the number of people with advanced heart failure is expected to rise by 30% by the year 2030. Some of the most common heart disease symptoms include chest pain, shortness of breath, and fatigue. It has been observed that machine learning can provide efficient heart disease prediction over a large amount of data thereby, empowering the patients and health care experts with the knowledge to make more accurate decisions at an appropriate time-bound. The central aspect of our proposed research is to build up a hybrid machine learning classifier using a Genetic Algorithm with a Super learner ensemble (GA-SLE) for the detection of heart disease with increased accuracy. In our study, the proposed system is compared with various machine learning algorithms such as Random Forest (RF), Multilayer Perceptron (MLP), K- Nearest Neighbour (KNN), Extra Tree (ET), Extreme Gradient Boosting (XGB), Super Vector Classifier (SVC), Stochastic Gradient Descent(SGD), AdaBoost, Decision Tree(CART), Gradient Boosting(GBM) using 10-Fold cross-validation and hyperparameter tuning. A regressed comparative analysis of all three approaches has been performed with the help of a table and plot. The proposed GA-SLE outperforms all other ML classifiers, obtaining a prediction accuracy of 99.8% with minimum error loss. As per our findings, the current GA-SLE classifier enables the practitioner to detect and diagnose the sickness in the early stage of the disease.