HEART DISEASE PREDICTION USING IMROVED-PSO AND ENSEMBLE CLASSIFIER

Abstract

Heart disease has recently become a leading cause of death worldwide, making the prediction of cardiovascular diseases a challenging task in clinical data analysis. With the rapid increase in the size and complexity of medical datasets, automated models leveraging data mining and machine learning techniques are proving essential for clinicians to make accurate and timely decisions. Key challenges in heart disease prediction include imbalanced datasets and the insufficient significance of certain features. The goal of this research is to propose effective feature optimization and classification methods to enhance heart disease prediction, facilitating early diagnosis for medical professionals. Initially, the min-max normalization technique is applied to rescale data from the Shahid Rajaei hospital and UCI Cleveland datasets. Then, optimal features and feature subsets are identified using the Mutual Information technique and the Improved Particle Swarm Optimization (IPSO) algorithm. Finally, the optimized feature subsets are fed into an ensemble classifier to predict the presence or absence of heart disease. Experimental results show that the proposed IPSO-based ensemble model achieved accuracy rates of 98.41% and 97.40% on the UCI Cleveland and Shahid Rajaei hospital datasets, respectively, outperforming traditional machine learning techniques.