A study on heart data analysis and prediction using advanced machine learning methods.

Cardiovascular diseases comprise a diverse array of disorders impacting the cardiac structure and vascular system and rank among the predominant factors contributing to mortality on a global scale. Every day, a significant number of individuals die from various heart-related issues. Therefore, early detection of heart diseases is of critical importance. Especially following these diagnoses, providing a more accurate diagnosis for individuals at high risk and subsequent extra treatments outline an essential roadmap for preventing heart attacks. This paper compares the performance of classic machine learning (ClassicML) and automated machine learning (AutoML) models across different variations that incorporate feature engineering and balancing techniques, thereby identifying which machine learning model is more successful in the development and implementation of patient-centered systems for the early prediction of cardiovascular diseases. The models used in this study utilize a combined dataset from Swiss, Hungarian, Cleveland, and Long Beach VA universities. This dataset consists of 14 main features, of which we used 12 features to analyze the individual experiencing a heart attack. The result of this classification problem is validated by accuracy. The accuracy result obtained is supported by the F1 score, precision, and recall results. The research encompasses nine traditional machine-learning algorithms as well as seven automated machine-learning algorithms. The findings of this investigation demonstrate that the performance of AutoML tools is not necessarily superior to traditional machine learning methodologies. Moreover, they highlight that effective feature extraction, when combined with an appropriate data balancing technique and a suitable machine learning model, can yield the best performance.