Enhanced binary grey wolf optimizer based on quantum computing and multi-strategy for feature selection on high-dimensional data classification

Feature selection is one of the major challenges in data mining and machine learning. The grey wolf optimizer (GWO) is a classical metaheuristic algorithm widely applied to various optimization problems due to its fast convergence speed and few parameters. However, GWO sometimes suffers from issues such as low convergence speed, insufficient population diversity, and a tendency to become trapped in local optima in the later stages of solving complex and specific optimization problems. To address these issues, this paper proposes a binary grey wolf optimizer based on quantum computing and multi-strategy enhancement (QMEbGWO), which is applied to feature selection in high-dimensional data classification. The innovations of this paper include an improved circular chaotic mapping method that combines quantum computing with a quantum gate mutation mechanism, a multi-population collaborative updating mechanism, and precise elimination and elastic generation strategies. Meanwhile, the continuous QMEbGWO is converted to its binary form using a V-shaped transfer function and a stochastic thresholding mechanism. Finally, to comprehensively evaluate the performance of QMEbGWO, we tested it on 21 high-dimensional datasets. The test results show that compared with eleven advanced feature selection methods, QMEbGWO’s average rankings in fitness value, feature subset size, accuracy, sensitivity, specificity, precision, MCC, and F1 Score are 3.79, 2.05, 4.72, 5.30, 5.25, 5.95, 5.72, and 5.68, respectively. In addition to the MCC final ranking second, the other in the first. These results demonstrate that QMEbGWO is an efficient and accurate feature selection method.