Scholarlink

<p>Cancer remains one of the most critical health challenges today. Feature selection is crucial in optimizing machine learning models by reducing dimensionality, improving efficiency, and enhancing classification accuracy. This paper proposes two parallel hybrid ensemble-based feature selection systems to improve computational efficiency and selection accuracy. The first system, Parallel Hybrid Ensemble SU-R Stepwise Search (PHE-SU-R-SS), iteratively refines feature subsets using the SU-R criterion. The second system, Parallel Hybrid Ensemble ChS-R Stepwise Search (PHE-ChS-R-SS)<b>,</b> employs the ChS-R approach for robust feature selection. A dynamic ensemble classifier integrating Random Forest (RF), Support Vector Machine (SVM), k-Nearest Neighbors (KNN), and Decision Tree (DT) is used for training and testing. Experimental results demonstrate that the proposed parallel methods achieve superior feature selection performance compared to conventional approaches. The parallel approach with ChS-R outperforms SU-R in ensemble ranking. Additionally, the parallel approach improves classification accuracy by 4–5% compared to the sequential approach without data partitioning. Parallel methods also enhance computational efficiency, achieving an average speedup of 4.74 over sequential methods. These findings highlight the effectiveness of parallel hybrid ensemble-based feature selection in optimizing machine learning models for high-dimensional biomedical data analysis.</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 Parallel Dynamic Weighted Ensemble Classifier for High Dimensional Cancer Datasets with Horizontal Partitioning

  • 期刊论文

  • 发表日期：

    2025年4月16日

查看全文
  • Archana Suhas Vaidya,
  • Dipak Patil,
  • Rahul Chakre

摘要

Cancer remains one of the most critical health challenges today. Feature selection is crucial in optimizing machine learning models by reducing dimensionality, improving efficiency, and enhancing classification accuracy. This paper proposes two parallel hybrid ensemble-based feature selection systems to improve computational efficiency and selection accuracy. The first system, Parallel Hybrid Ensemble SU-R Stepwise Search (PHE-SU-R-SS), iteratively refines feature subsets using the SU-R criterion. The second system, Parallel Hybrid Ensemble ChS-R Stepwise Search (PHE-ChS-R-SS), employs the ChS-R approach for robust feature selection. A dynamic ensemble classifier integrating Random Forest (RF), Support Vector Machine (SVM), k-Nearest Neighbors (KNN), and Decision Tree (DT) is used for training and testing. Experimental results demonstrate that the proposed parallel methods achieve superior feature selection performance compared to conventional approaches. The parallel approach with ChS-R outperforms SU-R in ensemble ranking. Additionally, the parallel approach improves classification accuracy by 4–5% compared to the sequential approach without data partitioning. Parallel methods also enhance computational efficiency, achieving an average speedup of 4.74 over sequential methods. These findings highlight the effectiveness of parallel hybrid ensemble-based feature selection in optimizing machine learning models for high-dimensional biomedical data analysis.