Scholarlink

<p>As one of the most prevalent causes of death globally, heart disease needs complex predictive models to support early detection and treatment. This study introduces a novel hybrid framework integrating Multi-Scale Feature Map Reconstruction with Inception in DenseNet (MSFMR-IncepDenseNet) for heart disease classification and an Analytical Hierarchy Process-VIKOR (AHP-VIKOR) feature selection model. The proposed framework incorporates comprehensive preprocessing steps, including mean imputation, min-max normalization, and DBSCAN-based data segmentation to ensure data integrity and effective clustering. AHP-VIKOR optimally ranks and selects significant features, enhancing model performance. The MSFMR-IncepDenseNet classifier leverages multi-scale attention mechanisms, ASPP modules, and Inception blocks within DenseNet, achieving superior classification accuracy through enriched feature representation. Results from experiments on typical datasets demonstrate that the suggested model performs noticeably better than current techniques including CNN, LSTM, XGBoost, and ANN, across multiple performance metrics like F1 score, specificity, sensitivity, and accuracy. The framework regularly attains an accuracy of 99.02%. in an 80:20 training-test split, showcasing its robustness in detecting heart disease. This hybrid strategy highlights how effective it may be as a tool for clinical settings, offering precise predictions to improve patient outcomes and inform timely medical decisions.</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

A hybrid framework for heart disease prediction using multi-scale feature map reconstruction with inception in DenseNet and AHP-VIKOR feature selection model

  • 期刊论文

  • 发表日期：

    2025年8月10日

查看全文
  • Raja Aswathi R,
  • K. Pazhanikumar

摘要

As one of the most prevalent causes of death globally, heart disease needs complex predictive models to support early detection and treatment. This study introduces a novel hybrid framework integrating Multi-Scale Feature Map Reconstruction with Inception in DenseNet (MSFMR-IncepDenseNet) for heart disease classification and an Analytical Hierarchy Process-VIKOR (AHP-VIKOR) feature selection model. The proposed framework incorporates comprehensive preprocessing steps, including mean imputation, min-max normalization, and DBSCAN-based data segmentation to ensure data integrity and effective clustering. AHP-VIKOR optimally ranks and selects significant features, enhancing model performance. The MSFMR-IncepDenseNet classifier leverages multi-scale attention mechanisms, ASPP modules, and Inception blocks within DenseNet, achieving superior classification accuracy through enriched feature representation. Results from experiments on typical datasets demonstrate that the suggested model performs noticeably better than current techniques including CNN, LSTM, XGBoost, and ANN, across multiple performance metrics like F1 score, specificity, sensitivity, and accuracy. The framework regularly attains an accuracy of 99.02%. in an 80:20 training-test split, showcasing its robustness in detecting heart disease. This hybrid strategy highlights how effective it may be as a tool for clinical settings, offering precise predictions to improve patient outcomes and inform timely medical decisions.