|Title||- A Multiobjective Genetic Algorithm for Feature Selection and Data Base Learning in Fuzzy-Rule Based Classification Systems|
|Publication Type||Book Chapter|
|Year of Publication||2003|
|Authors||Cordón, O., Herrera F., del Jesus M. J., Magdalena L., Sánchez A.M., and Villar P.|
|Editor||Bouchon-Meunier, Bernadette, Foulloy Laurent, and Yager Ronald R.|
|Book Title||Intelligent Systems for Information Processing|
|Pagination||315 - 326|
Publisher Summary This chapter illustrates a multiobjective genetic algorithm for feature selection and database learning in Fuzzy Rule-Based Classification System (FRBCS). An FRBCS presents two main components—the Inference System and the Knowledge Base (KB). The KB is composed of the Rule Base (RB) constituted by the collection of fuzzy rules, and of the Data Base (DB), containing the membership functions of the fuzzy partitions associated to the linguistic variables. The composition of the KB of an FRBCS directly depends on the problem being solved. If there is no expert information about the problem under solving, an automatic learning process must be used to derive the KB from examples. This contribution proposes a multiobjective genetic process for jointly performing feature selection and DB components learning that is combined with an efficient fuzzy classification rule generation method to obtain the complete KB for a descriptive FRBCS. This method achieves an important reduction of the relevant variables selected for the final system and adapts the fuzzy partition of each variable to the problem being solved. Therefore, the conclusion is that the proposed method allows for enhancing interpretability, accuracy, and performance of the FRBCS method.