Stroke remains a major global health burden and is a frequent and severe complication among patients with coronary heart disease (CHD). Early identification of individuals at high risk is essential for effective prevention; however, conventional clinical risk models often fail to capture complex nonlinear interactions among clinical and behavioral risk factors. This study aims to develop and evaluate an interpretable hybrid machine learning framework for accurate stroke risk prediction among individuals with CHD. A large real-world dataset comprising 253,680 participants was analyzed. Multiple machine learning algorithms were implemented, including logistic regression, tree-based ensembles, support vector machines, neural networks, and a stacked ensemble architecture. Class imbalance was addressed using stratified sampling combined with the Synthetic Minority Over-sampling Technique. Model performance was evaluated using discrimination , calibration, and scalability metrics, while explainable artificial intelligence techniques were applied to enhance clinical interpretability. The stacked ensemble demonstrated superior predictive performance, achieving an area under the receiver operating characteristic curve of 0.96, precision of 0.91, recall of 0.89, and a Matthews correlation coefficient of 0.82. LightGBM and XGBoost also exhibited strong discriminative ability, with AUC values of 0.94 and 0.95, respectively, while maintaining low computational latency. Explainability analyses identified 1 prior heart attack, body mass index, age, and lifestyle-related factors as key contributors to stroke risk. These findings demonstrate that hybrid ensemble learning combined with explainable artificial intelligence can substantially improve stroke risk prediction in CHD populations. The proposed framework provides clinically interpretable, scalable, and deployable decision support tools that can enhance early intervention strategies and support precision cardiovascular medicine.