Scholarlink

<p>Cardiovascular diseases (CVDs) remain the foremost cause of global mortality, accounting for nearly one-third of deaths worldwide. Accurate early prediction of CVD can assist clinicians in timely intervention and improved patient outcomes, especially in regions facing shortages of medical expertise. Computational intelligence techniques have shown promise in this area, yet challenges persist due to the high dimensionality of medical datasets, inter-feature correlations, and the difficulty of selecting informative variables. In this study, we introduce a Weighted Feature Matrix Method (WFMM) that systematically integrates multiple feature selection strategies to construct an optimal subset of attributes for CVD prediction. The approach assigns adaptive weights to features based on their frequency of selection across diverse algorithms, generating a transformed dataset representation that enhances model generalization. Twelve machine learning classifiers are then evaluated across three benchmark datasets (Cleveland, Statlog, and a combined UCI dataset) using accuracy, precision, recall, specificity, F1-score, and Matthews correlation coefficient as evaluation metrics. Experimental results demonstrate that WFMM consistently improves classification performance compared to baseline models using the full feature set. Notably, logistic regression and support vector classifiers achieve up to 4–9% accuracy gains after applying WFMM. The findings establish WFMM as a robust, classifier-agnostic framework for early CVD risk prediction, with potential for deployment as a decision-support tool in clinical environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Optimized Adaptive Feature Weighting Framework for Accurate Prediction of Cardiovascular Disease

  • 期刊论文

  • 发表日期：

    2026年1月27日

查看全文
  • Subham Kumar Padhy,
  • Anjali Mohapatra,
  • Sabyasachi Patra

摘要

Cardiovascular diseases (CVDs) remain the foremost cause of global mortality, accounting for nearly one-third of deaths worldwide. Accurate early prediction of CVD can assist clinicians in timely intervention and improved patient outcomes, especially in regions facing shortages of medical expertise. Computational intelligence techniques have shown promise in this area, yet challenges persist due to the high dimensionality of medical datasets, inter-feature correlations, and the difficulty of selecting informative variables. In this study, we introduce a Weighted Feature Matrix Method (WFMM) that systematically integrates multiple feature selection strategies to construct an optimal subset of attributes for CVD prediction. The approach assigns adaptive weights to features based on their frequency of selection across diverse algorithms, generating a transformed dataset representation that enhances model generalization. Twelve machine learning classifiers are then evaluated across three benchmark datasets (Cleveland, Statlog, and a combined UCI dataset) using accuracy, precision, recall, specificity, F1-score, and Matthews correlation coefficient as evaluation metrics. Experimental results demonstrate that WFMM consistently improves classification performance compared to baseline models using the full feature set. Notably, logistic regression and support vector classifiers achieve up to 4–9% accuracy gains after applying WFMM. The findings establish WFMM as a robust, classifier-agnostic framework for early CVD risk prediction, with potential for deployment as a decision-support tool in clinical environments.