<p>By removing irrelevant or redundant features, feature selection helps simplify the model and avoid over-fitting while preserving the core information of the data. In recent years, ensemble learning has attracted extensive attention in the field of feature selection because of its adaptability and diversity. Therefore, a feature selection method for an equilibrium optimizer based on the M-shaped transfer function family composed of integrated basic transfer functions is proposed. The integrated M-shaped transfer function family uses a novel ensemble learning framework to construct four M-shaped transfer function family models based on the existing basic transfer functions and realizes the evaluation and selection of feature subsets. Compared with the traditional innovative transfer function method, it can not only save a lot of researchers' energy and time but also make use of the idea of integration, it is easier to design a large number of new transfer function families with similar shapes in the existing transfer function interpretation. At the same time, in order to expand the new M-shaped transfer function set further, four MMr-shaped transfer functions are proposed on the basis of four M-shaped transfer functions using the reverse learning strategy. In addition, in order to further improve the efficiency of the feature selection model designed in this paper, the crossbar strategy and Lévy flight strategy are used to improve the equilibrium optimizer. The simulation experiment of the designed M-shaped transfer function group is carried out on the optimized equilibrium optimizer. Based on the performance of 8M-shaped and MMr transfer functions on 12 data sets, the best M-shaped transfer function is selected and compared with 8 basic transfer functions (two S-shaped, V-shaped, U-shaped, and RZ-shaped transfer functions each), to verify the validity of the M-shaped transfer function. Finally, the optimal M-shaped transfer function and the improved equilibrium optimizer are further compared with 6 advanced feature selection methods in recent years to verify the effectiveness and excellence of the improved strategy of M-shaped transfer function and algorithm. The experimental results show that the equilibrium optimizer based on the integrated M-shaped transfer function family has a good performance in solving feature selection problems, and can effectively improve the classification accuracy and obtain lower fitness values.</p>