International Journal of Applied Science and Engineering 2006. 4, 1: 41-52 Int. J. Appl. Sci. Eng., 2006. 4, 1 41 Generating Weighted Fuzzy Rules from Training Data for Dealing with the Iris Data Classification Problem Yung-Chou Chen a , Li-Hui Wang b , and Shyi-Ming Chen c* a Department of Electronic Engineering, National Taiwan University of Science and Technology, Taipei 106, Taiwan, R.O.C. b Department of Finance, Chihlee Institute of Technology, Banciao City, Taipei County 220, Taiwan, R.O.C. c Department of Computer Science and Information Engineering, National Taiwan University of Science and Technology, Taipei 106, Taiwan, R.O.C. Abstract: The most important task in the design of fuzzy classification systems is to find a set of fuzzy rules from training data to deal with a specific classification problem. In this paper, we present a new method to generate weighted fuzzy rules from training data to deal with the Iris data classification problem. First, we convert the training data to fuzzy rules, and then we merge those fuzzy rules in order to reduce the number of fuzzy rules. Then, we calculate the weight of each input variable appearing in the generated fuzzy rules by the relationships of input variables. The proposed weighted fuzzy rules generation method gets a higher average classification accu- racy rate than the existing methods. Keywords: fuzzy classification systems; fuzzy sets; Iris data; membership functions; weighted fuzzy rules. * Corresponding author; e-mail: smchen@et.ntust.edu.tw Accepted for Publication: August 17, 2005 © 2006 Chaoyang University of Technology, ISSN 1727-2394 1. Introduction One of important applications of fuzzy set theory [22] is in the fuzzy classification sys- tems. There are two approaches to obtain fuzzy rules for fuzzy classification systems. One of them is given directly by experts; the other is produced through an automatic learning process. In recent years, some meth- ods [1-10, 12-18, 20-21] have been presented to generate fuzzy rules from training in- stances. In [1], Castro et al. presented a method for learning maximal structure rules for dealing with the Iris data [11] classification problem. In [2], Castro et al. presented an inductive learning algorithm in fuzzy systems. In [3], Chang et al. presented a method to generate fuzzy rules from numerical data based on the exclusion of attribute terms for dealing with the Iris data classification problem. In [4], Chen et al. presented a method to generate fuzzy rules from numerical data for handling the Iris data classification problem. In [5], Chen et al. presented a method to generate fuzzy rules from relational database systems for estimating null values. In [6], Chen et al. presented a method for constructing fuzzy decision trees and generating fuzzy classifica- tion rules from training examples. In [9],