Scholarlink

<p>Efficiency and accuracy are two core metrics for evaluating the performance of a feature selection algorithm. Currently, most feature selection algorithms based on rough set (RS) theory struggle to balance these two critical indicators simultaneously. To address this challenge, we first investigate the RS theory under two technical frameworks: the Divisuin-Mining-Fusion (DMF) strategy and the Redundant-Sample-Deletion (RSD) technique. Theoretical analyses reveal that the DMF strategy and RSD technique can significantly improve the classification accuracy and computational efficiency of RS-based data analysis. Subsequently, we integrate the DMF strategy with the RSD technique to develop a novel feature selection algorithm. Results from comparative experiments confirm that, compared with traditional feature selection algorithms, the proposed algorithm exhibits significant advantages in both classification accuracy and computational efficiency.</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 feature selection algorithm embedded with division-mining-fusion strategy and redundant-sample-deletion technology

  • 期刊论文

  • 发表日期：

    2026年6月23日

查看全文
  • Qingzhao Kong,
  • Xinyu Tong,
  • Weihua Xu,
  • Jinpeng Hua,
  • Xiawei Zhang

摘要

Efficiency and accuracy are two core metrics for evaluating the performance of a feature selection algorithm. Currently, most feature selection algorithms based on rough set (RS) theory struggle to balance these two critical indicators simultaneously. To address this challenge, we first investigate the RS theory under two technical frameworks: the Divisuin-Mining-Fusion (DMF) strategy and the Redundant-Sample-Deletion (RSD) technique. Theoretical analyses reveal that the DMF strategy and RSD technique can significantly improve the classification accuracy and computational efficiency of RS-based data analysis. Subsequently, we integrate the DMF strategy with the RSD technique to develop a novel feature selection algorithm. Results from comparative experiments confirm that, compared with traditional feature selection algorithms, the proposed algorithm exhibits significant advantages in both classification accuracy and computational efficiency.