Interpretability of Heart Disease Detection Using a Robust Stacked Hybrid Model

Heart disease is a major global health issue, causing millions of deaths each year. Having a fast and reliable way to detect it can make a huge difference in saving lives and improving treatment outcomes. Our study introduces a stacked ensemble model that combines machine learning (ML) and deep learning to improve heart disease detection. The model is built using Gradient Boosting Machine (GBM), Extremely Randomized Trees (ET), and CatBoost (CB) as base classifiers, with an Artificial Neural Network (ANN) as the meta-classifier. To make sure the model's predictions are easy to understand, we used SHapley Additive exPlanations (SHAP) to break down how each feature contributes to the final decision. We trained and tested the model on a publicly available dataset from the UCI repository, one of the most widely used datasets for heart disease research. Metrics such as accuracy, precision, recall, F1-score, and AUC were used for evaluation, with the proposed model attaining an accuracy of 98.99%, surpassing state-of-the-art approaches. The results demonstrate the model's potential for real-world clinical application, offering both high accuracy and transparency in decision-making. Future work aims to further enhance generalizability and real-time adaptability for improved heart disease screening and diagnosis.