Cardiovascular Disease Detection through Innovative Imbalanced Learning and AUC Optimization

Document Type : Original Article


1 Associate Professor & Head, Department of Computer Science, Sri S.Ramasamy Naidu Memorial College, sattur, Virudhunagar, India

2 School of CS and IT, Jain Deemed-to-be University Bengaluru, Karnataka, India

3 Assistant Professor, Department of Computer Science, Sri S.Ramasamy Naidu Memorial College, Sattur, Virudhunagar, India


Cardiovascular diseases (CVDs) are a primary global health concern, impacting the heart and blood vessels extensively. In this paper, we introduce a novel imbalanced learning approach named Imbalanced Maximizing-Area Under the Curve (AUC) Proximal Support Vector Machine (ImAUC-PSVM), which harnesses the foundational principles of traditional PSVM for the detection of CVDs. The ImAUC-PSVM method offers several key advantages: 1) It skillfully incorporates AUC maximization directly into the objective function. This integration simplifies the model by reducing the number of parameters needing adjustment, making it particularly effective for handling imbalanced datasets through an efficient training process; 2) Theoretical analysis demonstrates that ImAUC-PSVM retains the same structural solution as standard PSVM. This similarity means it inherits PSVM's benefits, particularly in addressing progressive CVD scenarios with rapid incremental updates. Furthermore, we have incorporated a tailored Differential Evolution (DE) algorithm designed to navigate the complex hyperparameter space with finesse. The performance of this model was rigorously evaluated using comprehensive data from a medical survey conducted in 2012, which included an extensive cohort of 26,002 athletes. Critical parameters such as height, weight, age, gender, blood pressure, and resting heart rate were meticulously documented. The empirical results, benchmarked against established performance metrics, underscore the model's exceptional accuracy, solidifying its role as a reliable tool for CVD detection. This approach advances cardiovascular diagnostics and offers a scalable and adaptable solution, potentially influencing the broader landscape of healthcare analytics and patient care.