Effects of Feature Selection with 'Blurring' on NeuroFuzzy Systems
Abstract
Feature Selection is the problem of choosing a small subset of features that ideally is necessary and sufficient to describe the target concept. Feature selection is of paramount importance for any learning algorithm. We propose a new feature selection methodology based on the ‘Blurring’ measure, and empirically evaluate features selected through information-theoretic measures, stepwise multiple regression analyses, and the proposed method. We use neurofuzzy systems to compare the performance of these Feature Selection methods. Preliminary results using two data sets and the proposed Feature Selection method are promising.
Cite
Text
Piramuthu. "Effects of Feature Selection with 'Blurring' on NeuroFuzzy Systems." International Conference on Algorithmic Learning Theory, 1996. doi:10.1007/3-540-61863-5_41Markdown
[Piramuthu. "Effects of Feature Selection with 'Blurring' on NeuroFuzzy Systems." International Conference on Algorithmic Learning Theory, 1996.](https://mlanthology.org/alt/1996/piramuthu1996alt-effects/) doi:10.1007/3-540-61863-5_41BibTeX
@inproceedings{piramuthu1996alt-effects,
title = {{Effects of Feature Selection with 'Blurring' on NeuroFuzzy Systems}},
author = {Piramuthu, Selwyn},
booktitle = {International Conference on Algorithmic Learning Theory},
year = {1996},
pages = {135-142},
doi = {10.1007/3-540-61863-5_41},
url = {https://mlanthology.org/alt/1996/piramuthu1996alt-effects/}
}