UNSUPERVISED FEATURE-SELECTION USING A NEURO-FUZZY APPROACH

A neuro-fuzzy methodology is described which involves connectionist mi nimization of a fuzzy feature evaluation index with unsupervised train ing. The concept of a flexible membership function incorporating weigh ed distance is introduced in the evaluation index to make the modeling of clusters more appropriate. A set of optimal weighing coefficients in terms of networks parameters representing individual feature import ance is obtained through connectionist minimization. Besides, the inve stigation includes the development of another algorithm for ranking of different feature subsets using the aforesaid fuzzy evaluation index without neural networks. Results demonstrating the effectiveness of th e algorithms for various real life data are provided. (C) 1998 Publish ed by Elsevier Science B.V. All rights reserved.