Tradeoff between the performance of fuzzy rule-based classificationsystems and the number of fuzzy if-then rules

摘要

The main aim of the paper is to illustrate the tradeoff betweenthe performance of a fuzzy rule based classification system and its size(i.e., the number of fuzzy if-then rules) through computer simulationson commonly used data sets. In our computer simulations, we use a simpleheuristic method for generating fuzzy if-then rules from trainingpatterns, in which a pattern space is homogeneously partitioned intofuzzy subspaces by subdividing each axis into linguistic values. Forclearly illustrating the tradeoff, we use a genetic algorithm based ruleselection method with two objectives: to minimize the number of fuzzyif-then rules and to maximize the classification performance. Variousfuzzy rule based classification systems with different sizes aregenerated by the rule selection method for each data set