Scholarlink

<p>In heterogeneous data systems, the matrix-based attribute reduction is an important topic in rough set research, as it substantially enhances the performance of data classification and knowledge discovery. To date, numerous scholars have proposed diverse heuristic attribute reduction methods. However, most heuristic reduction iterations do not fully utilize the results of the previous iteration, resulting in repeated operations and affecting the efficiency of reduction. Therefore, inspired by incremental techniques, we propose a matrix-based fast reduction method to improve the efficiency of reduction. Firstly, we introduce the concepts of a cover relation matrix and a covering operator, and establish a zero-vector criterion. Then, we propose a reduction method that utilizes the cover relation matrix to solve positive regions. Secondly, we design two heuristic reduction algorithms, namely the forward heuristic reduction algorithm (FRCO) and the backward heuristic reduction algorithm (BRCO). On this basis, we adopt incremental computing strategies to dynamically update the relationship matrix to avoid repeated calculations and to speed up the reduction by gradually reducing the processing volume of samples to improve the efficiency of the algorithm, and then propose two improved reduction algorithms (QFRCO and QBRCO). Finally, we conduct comparative experiments with four classic feature selection methods on 12 UCI datasets to validate the effectiveness of our proposed method.</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 quick reduct method using covering operator in incomplete neighborhood rough sets

  • 期刊论文

  • 发表日期：

    2026年7月18日

查看全文
  • Hao Ge,
  • Bingcheng Li,
  • Chuanjian Yang,
  • Yi Xu,
  • Yuanting Yan

摘要

In heterogeneous data systems, the matrix-based attribute reduction is an important topic in rough set research, as it substantially enhances the performance of data classification and knowledge discovery. To date, numerous scholars have proposed diverse heuristic attribute reduction methods. However, most heuristic reduction iterations do not fully utilize the results of the previous iteration, resulting in repeated operations and affecting the efficiency of reduction. Therefore, inspired by incremental techniques, we propose a matrix-based fast reduction method to improve the efficiency of reduction. Firstly, we introduce the concepts of a cover relation matrix and a covering operator, and establish a zero-vector criterion. Then, we propose a reduction method that utilizes the cover relation matrix to solve positive regions. Secondly, we design two heuristic reduction algorithms, namely the forward heuristic reduction algorithm (FRCO) and the backward heuristic reduction algorithm (BRCO). On this basis, we adopt incremental computing strategies to dynamically update the relationship matrix to avoid repeated calculations and to speed up the reduction by gradually reducing the processing volume of samples to improve the efficiency of the algorithm, and then propose two improved reduction algorithms (QFRCO and QBRCO). Finally, we conduct comparative experiments with four classic feature selection methods on 12 UCI datasets to validate the effectiveness of our proposed method.