An improved evolutionary wrapper-filter feature selection approach with a new initialisation scheme

Treated as one of the popular measures in information theory, fuzzy mutual quantifies amount that random variable has about another one. Different from standard information, can deal with not only discrete-valued but also real-valued variables. Therefore, been recently used evolutionary filter feature selection approaches to measure correlation between classes and features, dependencies within a set. Typically, this way be considered computationally efficient sometimes it may contribute performance classification algorithm. To address issue, an improved wrapper-filter approach which integrates initialisation scheme local search module based on differential evolution is proposed. According number experiments conducted several real-world benchmark datasets, proposed does significantly improve computational efficiency computation technique