Reening the Rules Obtained by Slave Reening the Rules Obtained by Slave

In this work, we describe SLAVE+R, a fuzzy rule learning system based on SLAVE that includes a reenement module. SLAVE (Structural Learning Algorithm in Vague Environment) was developed for working with noise-aaected systems where the application of some conditions of classical learning theory do not produce good descriptions. This learning system allows to obtain the structure of the rule, i.e., it can determine among all the variables proposed those that are relevant for describing the system (feature selection). SLAVE uses an iterative approach for learning with genetic algorithms. This method is an alternative approach from the classical Pittsburgh and Michigan approach and it consists of obtaining in each iteration an useful rule for describing the system. So, in this approach, the nal solution is obtained from partial solutions. The reenement module appened to SLAVE (SLAVE+R) is proposed as a method for verifying that the union of the partial solutions is a good global solution. Furthermore, this module allows to minimize the number of the necessary rules, keeping the accuracy and to improve the comprensibility of these rules.