Evolutionary Approach to Feature Selection with Associative Models
Angel Ferreira Santiago, Cornelio Yáñez-Márquéz, Mario Aldape-Pérez, Itzamá López-Yáñez · Research in Computing Science · 2014
Feature selection aims to nd ways to single out the subset of features which best represents the phenomenon at hand and improves performance.This paper presents an approach based on evolutionary computation and the associative paradigm for classication.A wrapperstyle search guided by a genetic algorithm uses the Hybrid Associative Classier to evaluate candidate solutions and thus approximate the optimal feature subset for dierent data sets.The results suggest that this is a feasible approach for feature selection, obtaining solutions equal or similar to the optimal solution while evaluating a relatively small fraction of the search space.