Unveiling Key Predictors for Early Heart Attack Detection using Machine Learning and Explainable AI Technique with LIME

Abstract

The prominence of cardiovascular diseases, particularly heart attacks, as a leading cause of global mortality is highlighted, with an increasing number of deaths attributed to cardiovascular diseases over the years. Amidst these challenges, artificial intelligence (AI) and machine learning (ML) technologies emerge as powerful tools in healthcare. This study conducts a comparative analysis of predictive features extracted from diverse classification algorithms, including AdaBoost Classifier (ABC), Random Forest (RF), Gradient Boosting Classifier(GBC) and Light Gradient-Boosting Machine (LGBM), aiming to identify common patterns in predictive outcomes. LGBM emerges as the standout performer among classification algorithms, boasting a remarkable average training accuracy of 99.33%. Results demonstrate comparable precision, recall, and F1 scores among RF, GB, and LGBM, while ABC lags behind. The study reveals from eXplainable AI technique that consistent attribution of importance to attributes like "kcm" and "troponin" across all methods for classifying "Attack" instances, indicating their pivotal role in prediction. The research underscores the potential clinical application of machine learning for heart attack diagnosis and suggests the adoption of various deep learning techniques to enhance predictive performance.