The relationship between the number of features and classification accuracy in the feature selection can usually be considered as a multimodal multi-objective optimization problem (MMOP). However, most existing multimodal multi-objective evolutionary algorithms (MMOEAs) have weak search capability in solving high-dimensional MMOPs. To solve this problem, a two-stage feature selection method based on causal model and MMOEA is proposed. In the proposed algorithm, the causal model is firstly used to select features to reduce the problem dimensionality, and then a competitive MMOEA is used to find a subset of features with multimodal characteristics. To demonstrate the performance of the proposed algorithm, it is used to solve the feature selection problem of intraoperative hypothermia prediction model. Experimental results show that the proposed two-stage feature selection method not only combines the advantages of two different methods, but also provides more decision support for the intraoperative hypothermia prediction.

causal model; multimodal multi-objective optimization; feature selection; intraoperative hypothermia; evolutionary computation