Soft computing techniques for intelligent classification system: a case study

Soft computing techniques are becoming popular in designing real world industrial applications. Researchers are trying to integrate different soft computing paradigms such as fuzzy logic, artificial neural network, genetic algorithms etc., to develop hybrid intelligent autonomous systems that provide more flexibility by exploiting tolerance and uncertainty of real life situations. Intelligent classification systems are the most well known attempts. In this work a neuro fuzzy feature selector has been designed which is capable of extracting information in the form of fuzzy rules from numeric as well as non-numeric (linguistic) data. Conventional MLP and a variation of it have been used as the neural models and their performance has been compared by simulation with two different data sets. It is found that the proposed variation of the conventional MLP is better in respect to training time and classification accuracy.