Hybrid Attention-Driven Deep Learning Framework for Cardiovascular Disease Classification

Cardiovascular diseases (CVDs) remain the leading cause of global mortality, underscoring the urgent need for early and accurate detection methods that surpass the limitations of conventional diagnostic approaches. This study proposes a Hybrid Attention-Driven Deep Learning Framework designed to enhance predictive accuracy by effectively leveraging heterogeneous clinical features. The dataset, obtained from the UCI repository, comprises 918 patient records containing demographic, physiological, and electrocardiographic attributes relevant to heart disease risk assessment. The proposed model integrates a dual-branch architecture one dedicated to processing numerical features through dense transformations and the other to encoding categorical variables using embeddings followed by a Multi-Head Attention Fusion Layer that captures complex inter-feature dependencies. Experimental results demonstrate that the proposed framework outperforms several traditional machine learning models, achieving a test accuracy of 0.897, an F1-score of 0.910, and an AUC of 0.930. These findings highlight the model’s robustness, superior generalization capability, and strong potential for clinical decision support in early CVD diagnosis.