Opposition-based moth-flame optimization improved by differential evolution for feature selection

This paper provides an alternative method for creating an optimal subset from features which in turn represent the whole features through improving the moth-flame optimization (MFO) efficiency in searching for such optimal subset. The improvement is performed by combining the opposition-based learning technique and the differential evolution approach with the MFO. The opposition-based learning is used to generate an optimal initial population to improve the convergence of the MFO; meanwhile, the differential evolution is applied to improve the exploitation ability of the MFO. Therefore, the proposed method noted as OMFODE has the ability to avoid getting stuck in a local optimal value, unlike the traditional MFO algorithm and increase the fast convergence. The performance evaluation of our approach will be through a group of experimental results. In the first one, the proposed method has been tested over several CEC2005 benchmark functions. The second experimental series aims to assess the quality of the proposed method to improve the classification of ten UCI datasets by performing feature selection on such datasets. Another experiment is testing our method for classifying a real dataset, which represents some types of the galaxy images. The experimental results illustrated that the proposed algorithm is superior to the state-of-the-art meta-heuristic algorithms in terms of the performance measures.