Abstract:
The development of precised, effective, and non-invasive diagnostic process is necessary for early intervention since coronary artery disease (CAD) continues to be a major worldwide health concernd. Traditional machine learning models have shown promised in predicting healthcare, but because they are good and difficult to explain, their "black box" character presents a major obstacled to practical application. By combining explaining artificial intelligence (XAI) method with a high-performancess ensemble learning methodology, this study offers a novel framework that tackles this pressing issue. The proposed solution incorporates extremely cautiously the predictived natured of LightGBM and XGBoost by meaning of a voting ensemble design as their hyperparameters are optimized with the aid of Bayesian models. The aim behind the hybrid architecture is to deliver highs accuracy of forecast as well as trends of complexity data. The effectivenes of the models was evaluated comprehensively using a variety of measures, including accuracy, precision, recall and the under the receiver operating characteristic curve (AUC). Based on the experimentals findings, the based classifiers approach is significantly better than the voting ensemble model with an 96.218 test accuracy. The analysis funds on SHapley Additive exPlanations (SHAP) that involves the decomposition of the opaque decision-making process in the model to obtain comprehensible and practical information. The three features that are most frequently highlighted in the forecasting of CAD by the XAI analysis are cholesterol and resting blood pressure in addition to the nature of the chest palpation. The SHAP analysis provides a deeper view on their positive or negative relationship with disease risk. This thesis demonstrated the important of improved predictive performancd between highly developed ensemble modeling and XAI as well as contributed to the interpretability and trust required to facilitated the seamless integration of AI-oriented systems into a clinical decision-making-driven environment
