Heart Failure Prediction Using Machine Learning with Metaheuristic Feature Selection Techniques

Abstract

Heart failure is a major complication of cardiovascular diseases (CVDs), which are the leading cause of mortality globally. However, early treatment and detection of heart failure may increase the chance of survival. Using current clinical data, machine learning (ML) algorithms present an effective approach for predicting heart failure. Utilizing ML algorithms and feature selection using metaheuristic methods, we present a novel framework in this research for the prediction of heart failure. We employ several ML algorithms and perform feature selection using four meta-heuristic techniques: Grey Wolf Optimization (GWO), Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and Ant Colony Optimization (ACO). The performance of each combination is evaluated and compared based on the F-score metric, which we aim to maximize. (PSO) has achieved the highest when choosing the relevant features which increased the overall accuracy from 0.83 to 0.90. The results indicate that our proposed framework can effectively identify relevant features and improve the predictive performance of ML algorithms for heart failure. Furthermore, we provide a comprehensive comparison of the meta-heuristic techniques, highlighting their advantages and limitations for feature selection in heart failure prediction.