Unsupervised neuro-fuzzy feature selection

This article describes a neuro-fuzzy methodology for feature selection under unsupervised training. The methodology includes connectionist minimization of a fuzzy feature evaluation index. A concept of flexible membership function incorporating weighted distance is introduced in the evaluation index to make the modeling of clusters more appropriate. A set of optimal weighting coefficients in terms of networks parameters representing individual feature importance is obtained through connectionist minimization. Besides this, another algorithm is developed for ranking different feature subsets using the fuzzy evaluation index without neural networks. Results demonstrating the effectiveness of the algorithms for various real life data are provided