Binary Multi-objective Hybrid Equilibrium Optimizer Algorithm for Microarray Data

Feature selection aims at identifying features relevant to the target from high-dimensional data to enhance the performance of the learner. When dealing with high-dimensional data, traditional methods often exhibit lower accuracy, posing significant challenges to feature selection. In this paper, we propose a multi-objective hybrid binary balance optimizer algorithm to address this issue. This method integrates the output results of multiple filters, considering redundancy and complementarity among genes in the process. Building upon the original Equilibrium optimizer (EO), we employ an external archive to guide the population’s search direction and discretize the EO using an S-shaped transfer function. In this approach, we do not consider the classification error rate as a sole optimization objective. Instead, we use the ranking of genes in the selected subset across various filters as the second optimization objective. This design allows for the selection of a smaller number of genes while capturing those most relevant to the labels. To validate the performance of the proposed method, we compare it with 5 multi-objective algorithms on 12 microarray datasets and 3 UCL datasets. The results demonstrate that the proposed method consistently achieves the optimal Pareto front on most datasets.