计算机应用研究Issue(5)：1305-1308,4.DOI:10.3969/j.issn.1001-3695.2015.05.006
一种基于特征聚类的特征选择方法
Novel feature selection method based on feature clustering
摘要
Abstract
Feature selection has become a very useful pre-processing technology in data mining and machine learning.This paper proposed a mean-similarity measure and a new feature selection method based on feature clustering (named FSFC)in the unsupervised learning.Firstly,the method divided the entire feature space into a set of homogeneous subspaces when a clustering algorithm was used for the full feature set.Then it formed the final feature set by selecting some representative fea-tures from each cluster.At last,it removed the irrelevant and redundant features.Experimental results on UCI datasets show that the performance of dimensionality reduction and classification with C4.5 and naive Bayes obtained by FSFC is close to the several states of art supervised feature selection algorithms.关键词
特征选择/特征聚类/相关度/无监督学习Key words
feature selection/feature clustering/similarity/unsupervised learning分类
信息技术与安全科学引用本文复制引用
王连喜,蒋盛益..一种基于特征聚类的特征选择方法[J].计算机应用研究,2015,(5):1305-1308,4.基金项目
国家自然科学基金资助项目（61202271）；国家社会科学基金资助项目（13CGL130）；国家教育部人文社会科学资助项目（14YJC870021）；广东省自然科学基金资助项目（S2012040007184）；广东省普通高校科技创新资助项目（2012KJCX0049，2013KJCX0069）；广东省科技计划资助项目 （）
