WJMI: A New Feature Selection Algorithm Based on Weighted Joint Mutual Information

DOI

10.2991/icmii-15.2015.108How to use a DOI?

Keywords

feature selection, feature interaction, pruning rule, WJMI

Abstract

Mutual information based feature selection algorithms are very popular currently. Though they perform well in many cases, they suffer from two drawbacks: (1) the neglect of feature interaction; (2) the overestimation of some features. To overcome these shortcomings, a new feature evaluation criterion considering feature interaction is proposed and a pruning rule is designed. Based on the criterion and pruning rule, a new feature selection algorithm WJMI is proposed. Experiments carried out on UCI real world dataset against other four algorithms demonstrate the effectiveness of WJMI.

Copyright

© 2015, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Xiuli Qi
AU  - Chengxiang Yin
AU  - Kai Cheng
AU  - Xianglin Liao
AU  - Xingdang Kang
PY  - 2015/10
DA  - 2015/10
TI  - WJMI: A New Feature Selection Algorithm Based on Weighted Joint Mutual Information
BT  - Proceedings of the 3rd International Conference on Mechatronics and Industrial Informatics
PB  - Atlantis Press
SP  - 632
EP  - 638
SN  - 2352-538X
UR  - https://doi.org/10.2991/icmii-15.2015.108
DO  - 10.2991/icmii-15.2015.108
ID  - Qi2015/10
ER  -