Compensatory Fuzzy Logic Genetic Algorithm for Classification Problems: A Case Study

Abstract

This article presents an approach for creating fuzzy predicates using genetic algorithms. The proposed method incorporates an internal genetic algorithm to optimize the membership functions of the linguistic variables involved in the discovered predicates, taking advantage of statistical data for the initialization of the population and taboo and weighted roulettes for the construction of the predicates. The generation offuzzy predicates is based on the implication and equivalence operators, as well as on deductive structures, such as modus ponens. Furthermore, the evaluation of predicates on data sets is based on For All and Exists quantifier operators, which also guide the search for the best predicates according to their truth values. Furthermore, the popular Iris database is used asa case study to demonstrate the effectiveness and applicability of this approach.