摘要： 随着特征维数的不断增长，搜索性特征选择算法付出巨大的时间代价，而非搜索性算法则由于其时间代价小，而且能有效去除冗余特征等优越性越来越受到关注。该文介绍了一种非搜索性算法——KNN特征选择算法，该算法通过计算特征间的相关性来消除冗余特征，时间代价小。在此基础上，该文提出了一种基于KNN算法的组合式非搜索特征选择算法。

关键词: 特征选择, 非搜索, 最大信息压缩指数, KNN

Abstract: As feature dimension increases continually, search algorithms costs a lot of time. Therefore, much attention is paid to no-search method owing to their small time cost and efficiency in reducing redundant features. KNN method, which is based on measuring similarity between features whereby redundancy therein is removed is introduced. And a no-search combined feature selection method based on KNN is proposed.

Key words: feature selection, no-search, maximum information compression index, KNN

中图分类号: 

• TP391.4