Filter-Wrapper Hybrid Method on Feature Selection

Feature selection is a process commonly used in machine learning. This paper examines two broad classes of feature selection methods: filter methods and wrapper methods to find their individual advantages and disadvantages. This paper selects their different merits to propose a filter-Wrapper hybrid method (FWHM) to optimize the efficiency of feature selection. FWHM is divided into two phase, which orders these features according to a reasonable criterion at first, then select best features based on final criterion. These experiments on benchmark model and engineering model prove that FWHM has better performances both in accuracy and efficiency more than conventional methods.