分布优势数据集的矩阵增量属性约简算法  

A Matrix-based Incremental Reduction Approach for Distributed Dominant Data Set

作　　者：景运革 吉新如 王鹏岭 王宝丽 JING Yunge;JI Xinru;WANG Pengling;WANG Baoli(School of Mathematics and Information Technology,Yuncheng University,Yuncheng 044000,China)

机构地区：[1]运城学院数学与信息技术学院,山西运城044000

出　　处：《山西大学学报(自然科学版)》2025年第1期158-168,共11页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(61703363);山西省基础研究计划项目(201801D121148);运城学院数据挖掘与工业智能应用科研创新团队(YCXYTD-202402);运城学院科研项目(321701)。

摘　　要：传统属性约简算法不能有效解决动态数据属性约简问题,寻求高效动态数据属性约简算法是目前人工智能领域研究的热点。本文在动态分布优势数据集中引入矩阵优势条件熵和优势矩阵,探讨基于优势条件熵的矩阵增量属性约简方法。首先,定义了分布数据集的优势矩阵和优势条件熵;其次,通过分析分布数据集添加对象的过程,提出了优势矩阵的增量更新原理和融合机制;然后,给出了基于优势条件熵的矩阵增量约简方法。最后,利用6组UCI(University of California Irvine)优势数据集进行实验,用于验证增量属性约简算法的高效性。实验结果表明:与非增量属性约简算法相比,由增量属性约简算法计算约简的运行时间缩短了85.6%。所以,本文所给出的矩阵增量属性约简算法是求解动态分布优势数据集属性约简的快速有效方法。Traditional attribute reduction algorithms are inefficient to deal with dynamic decision systems,and seeking efficient reduction approach of dynamic data is a research hotspot in the field of artificial intelligence.The paper studies dominant conditional entropy-based incremental attribute reduction approach by introducing dominant conditional entropy and dominant matrix into the dynamic distributed dominant datasets.Firstly,the dominant matrix and dominant conditional entropy of distributed dominant data set are defined.Secondly,an incremental learning mechanism and fusion mechanism of dominant matrix are proposed by analyzing the process of the adding some objects into the distributed dominant data set.Then,a matrix-based incremental attribute reduction approach based on dominant conditional entropy is presented.Finally,experiments on six UCI datasets were conducted to validate the efficiency of the incremental attribute reduction approach.The experimental results showed that the reduction time of incremental attribute reduction approach was reduced by an average of 85.6%compared with the non-incremental reduction approach.Therefore,the proposed matrix-based incremental attribute reduction method is quick and effective in solving reduction of dynamic distributed dominant datasets.

关 键 词：分布优势数据集 增量技术 属性约简 优势条件熵 优势数据集矩阵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]