COMPARATIVE STUDY OF MACHINE LEARNING MODELS FOR CARDIOVASCULAR DISEASE PREDICTION

Abstract

Cardiovascular disease (CVD) remains one of the leading causes of death globally, necessitating timely and accurate prediction tools for early diagnosis and intervention. Recent advancements in machine learning (ML) have shown significant promise in enhancing predictive models for CVD by leveraging complex patterns in medical data. This comparative study critically examines multiple peer-reviewed research papers that utilize ML techniques for CVD prediction. The analysis focuses on research metho-dologies, preprocessing techniques, dataset characteristics, and algorithms applied, performance out-comes, and reported limitations. A wide range of ML models—including ensemble methods like XGBoost and Rotation Forest, as well as traditional classifiers such as KNN and Logistic Regression—were employed across studies. Results indicate that ensemble models often outperform others in terms of accuracy and F1-score, with some models achieving up to 98.50% accuracy. However, issues such as limited dataset diversity, lack of real-time validation, and absence of multimodal integration constrain broader applicability. This paper underscores the importance of using diverse and large-scale datasets, real-time data acquisition, and cross-validation to improve model generalizability. By highlighting both the strengths and gaps in existing research, this study offers insights into future directions for developing robust, clinically viable ML-based CVD prediction systems.