计算机应用研究2016，Vol.33Issue(5)：1322-1326,5.DOI:10.3969/j.issn.1001-3695.2016.05.010
一种基于组策略的过滤式特征选择算法
Filter feature selection algorithm based on group mechanism
摘要
Abstract
The MRMR algorithm was fast and effective,and it was popular in the handling of high-dimension data.Motivated by this,this paper proposed a refined MRMR algorithm (MRMRE)based on the group mechanism.To improve the results of feature selection,this algorithm not only considered the relationship between features,but also considered the relationship be-tween feature groups.This algorithm took the MRMR algorithm as the frame,used CCA as the measure and selected SVMs as the base classifier.Massive experiments conducted on generous images and gene sequence data sets in the machine learning da-tabase from UCI show that the proposed algorithm has higher result stability and classification precision in feature selection compared to the MRMR algorithm.关键词
特征选择/组策略/最大相关最小冗余算法/典型相关分析Key words
feature selection/group mechanism/MRMR(max-relevance and min-redundancy)/CCA(canonical correla-tion analysis)分类
信息技术与安全科学引用本文复制引用
许尧,胡学钢,李培培..一种基于组策略的过滤式特征选择算法[J].计算机应用研究,2016,33(5):1322-1326,5.基金项目
国家自然科学基金资助项目（61273292）；国家教育部博士点博导基金资助项目 （）
