Learning Linguistic Fuzzy Rules by Using Estimation of Distribution Algorithms as the Search Engine in the COR Methodology

Learning models from data which have the double ability of being predictive and descriptive at the same time is currently one of the major goals of machine learning and data mining. Linguistic (or descriptive) fuzzy rule-based systems possess a good tradeoff between the aforementioned features and thus have received increasing attention in the last few years. In this chapter we propose the use of estimation of distribution algorithms (EDAs) to guide the search of a good linguistic fuzzy rule system. To do this, we integrate EDAs in a recent methodology (COR) which tries to take advantage of the cooperation among rules. Experiments are carried out with univariate and bivariate EDAs over four test functions, and the results show that the exploitation of (pairwise) dependencies done by bivariate EDAs yield to a better performance than univariate EDAs or genetic algorithms.