Review on wrapper feature selection approaches

The main objective of feature selection process consists of investigating the optimal feature subset leading to better classification quality while spending less computational cost compared to maintaining the whole initial feature set. The problem of feature selection has been extensively researched since the early beginnings of machine learning. Even though several methods were proposed to handle the issue of feature selection using a variety of techniques, it is difficult to identify a specific method as the most fitted one regarding the feature subset selection issue. In this study, we provide an overview of existing wrapper methods pointing out their weaknesses and their strengths.