Generating Fuzzy Rules from Training Instances for Fuzzy Classification Systems

Document Type : Original Article

Authors

1 Assistant Professor, Ferdowsi University of Mashhad.

2 Professor, Tehran University.

3 Assistant Professor, Tehran University.

Abstract

In recent years many methods to generate fuzzy rules based on training samples is proposed. One of the methods that recently suggested is Chen & Tsai’s method that was presented in 2008. This method was based on three indices: Attribute Threshold Value ( ), Classification Threshold Value ( ) and Level Threshold Value ( ). This method has a higher average classification accuracy rate than other methods available. In part of generating interim rules, Authors have provided suggestions to improve method and they used modified method in a case study. What in this research has been considered is the examining of CT08 improved method in an applied field on the one hand, and presentation of a classification system for mobile sets on the other hand. Improved method depicted higher accuracy than original method (about 0.2 in mobile set subject).

Keywords