Development of a Hyperparameter-Optimized Decision Support System for Cardiovascular Disease Prediction

Abstract

Cardiovascular disease (CVD) remains one of the leading health dilemmas all over the world, and it is among the leading causes of morbidity and mortality. To solve this problem, more than traditional diagnostic methods are needed; smart decision support systems (DSS) are required to help clinicians recognize the condition at an early stage and provide advice on treatment regimens. We develop an improved DSS with multiple machine learning (ML) techniques in cardiovascular risk prediction. Our framework is also focused on clinical applications, as compared to many other prior models that are seen as only theoretically viable. To determine its adequacy, eight ML and deep learning (DL) models were trained and optimized on a clinical set founded on feature selection and hyper-parameter tuning schemes. Among these, the XGBoost classifier exceeded by far the others in terms of accuracy, interpretability, and speed of computation and its operation, and as such would be the best candidate to deploy. Another characteristic of our system is that the Shapley Additive Explanations (SHAP) analysis is applied, which facilitates increasing the confidence of the results by clearly indicating how they may be compiled for clinicians. The benefits of the proposed DSS lie not only in supporting accurate diagnosis but also in translating to real-time reports and recommendations, which are both actionable and supportive of patient management. Additionally, its architecture is scalable and can fit in a variety of healthcare systems and help address the issue of early intervention to reduce the burden of CVD as a whole. Received: 5 February 2025 | Revised: 18 August 2025 | Accepted: 30 August 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement The data that support the findings of this study are openly available in the CardiovascularDataset repository at https://github.com/zubi00/CardioviscularDataset. Author Contribution Statement Uzma Nawaz: Conceptualization, Methodology, Software, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Project administration. Mufti Anees-ur-Rahaman: Methodology, Software, Writing – review & editing. Hafiz Muhammad Ubaidullah: Software validation, Formal analysis, Writing – review & editing. Chaudhry Muhammad Ali Nawaz: Visualization, Investigation, Resources. Zubair Saeed: Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision.

Ähnliche Arbeiten