Coronary Artery Disease (CAD), a leading cause of death worldwide, requires prompt and accurate diagnosis to enable effective treatment and avoid unnecessary or invasive procedures. Traditional statistical methods, such as those proposed by the European Society of Cardiology (ESC), often fall short in delivering reliable predictive performance, particularly in heterogeneous patient populations. To address this gap, this paper examines the application of Machine Learning (ML) to enhance CAD risk assessment and facilitate informed diagnostic decision-making. We propose a stacking ensemble model that combines multiple classifiers and achieves an accuracy of 83%, outperforming individual models. Building on this model, we introduce CARDiA, a human-centered, intelligent clinical decision support system (CDSS) designed to assist cardiologists in evaluating CAD likelihood and determining appropriate diagnostic actions. Beyond prediction, CARDiA integrates ESC clinical guidelines with a workflow-based reasoning engine to provide interpretable, guideline-aligned diagnostic pathways. By combining predictive accuracy with evidence-based, process-driven recommendations, CARDiA offers a robust and user-friendly tool to enhance CAD diagnosis in clinical practice.