Unveiling the Contributing Features in Heart Disease Occurrence: An Explainable AI Approach

Heart disease is a leading cause of mortality worldwide. Accurate prediction of heart disease and understanding the contributing factors for heart disease are crucial. Recently, Explainable AI (XAI) techniques have been utilized to unravel the underlying factors associated with heart disease. This research paper primarily focuses on identifying the significant features for occurrence of heart disease. Firstly, data preprocessing is performed, followed by feature selection using the chi-square test for categorical features and the ANOVA test for numerical features. Five Machine Learning (ML) models, namely Logistic Regression (LR), Random Forest (RF), Naive Bayes (NB), Gradient Boosting (GB), and Adaptive Boosting (AB), are then employed for heart disease prediction. Among these models, the LR model demonstrates the highest accuracy of 87.02%, outperforming the others. The LR model, which achieves the best performance, is further analyzed using XAI techniques, including Permutation Importance, SHAP (SHapley Additive exPlanations), and LIME (Local Interpretable Model-agnostic Explanations). These techniques reveal that the ST _slope feature consistently emerges as the most important feature for heart disease. Finally, visual representations are employed to showcase the importance of features in occurring heart disease. This research work sheds light on the application of XAI techniques in identifying important features for heart disease occurrence and provides valuable insights for further investigation in this domain.