Predicting Heart Diseases Using Machine Learning and Different Data Classification Techniques

Abstract

Heart disease is a leading cause of mortality worldwide, with early detection playing a critical role inreducing death rates. Accurate prediction of heart disease remains challenging due to complex medical data andthe inability to provide continuous monitoring. Utilizing the Heart Disease dataset, various feature selectiontechniques, including ANOVA F-statistic (ANOVA FS), Chi-squared test (Chi2 FS), and Mutual Information (MIFS), were employed to identify significant predictors. Synthetic Minority Oversampling Technique (SMOTE)was applied to address data imbalance and enhance model performance. A comprehensive classification approachwas undertaken using diverse machine learning models and ensemble methods. Among these, a StackingClassifier combining Boosted Decision Trees, Extra Trees, and Light GBM achieved superior results, delivering 100% accuracy across all feature selection techniques. The high performance highlights the effectiveness ofadvanced ensemble learning in achieving reliable heart disease predictions, emphasizing the potential ofintegrating robust feature selection with sophisticated classification models for precise medical data analysis. Thisapproach demonstrates the capacity to support early diagnosis and improved patient outcomes.