Résumé: In most data mining tasks, feature selection (FS) is a necessary preprocessing step that can reduce the dimensionality of the dataset while ensuring adequate classification accuracy. In this paper, a ReliefF-guided novel binary equilibrium optimizer (RG-NBEO) is proposed for feature selection. Based on the binary equilibrium optimizer, two novel mechanisms are employed to improve the evolution performance. First, two novel transfer functions (SSr and VVr) based on the concept of opposition learning are proposed to transform the continuous search space into a binary search space and achieve a good balance between exploration and exploitation. Second, a ReliefF bootstrapping strategy is proposed to add and remove features directionally in the iterative process according to the feature weights. The simulation experiments are first based on the equilibrium optimizer (EO) variants constructed from the classical S- and V-shaped transfer functions. The variant EO with the best performance is selected and compared with five superior swarm intelligence optimization algorithms and six classical filter feature selection algorithms. The performance of the proposed method was tested on 18 standard datasets, and the results of the different algorithms were statistically evaluated using the Wilcoxon rank sum test and the Freidman rank sum test. The results show that this method can effectively improve the classification accuracy in most cases.

Langue: Anglais