PERFORMANCE ANALYSIS OF MACHINE LEARNING APPROACHES IN HEART DISEASE PREDICTION

Abstract

Cardiovascular diseases (CVDs) are among the leading causes of mortality worldwide, plac-ing increasing pressure on healthcare systems to adopt advanced techniques for early detec-tion. Machine learning (ML) has emerged as a powerful approach to analyze patient data and predict cardiovascular risk with high accuracy. The availability of structured datasets and modern computational techniques has enabled researchers to compare and optimize ML algo-rithms for effective diagnosis. This paper presents a comprehensive analysis of various ma-chine learning approaches used in recent studies on heart disease prediction. Using a summa-rized dataset of five research works, the paper evaluates methodologies, preprocessing strate-gies, datasets, and performance metrics. Results indicate that ensemble models, particularly XGBoost, achieve superior performance, with up to 98.5% accuracy. Hybrid and tree-based models also demonstrate strong predictive capabilities. The paper highlights existing gaps, such as limited dataset diversity, lack of multimodal integration, and insufficient focus on explainability. Future directions include incorporating advanced deep learning architectures, explainable AI (XAI), multimodal fusion, and real-time predictive systems. The purpose of this analysis is to guide researchers toward more robust and clinically applicable ML-based heart disease prediction frameworks.