基于因果模型和多模态多目标优化的两阶段特征选择方法  

A two-stage feature selection method based on causal model and multimodal multi-objective optimization

作　　者：王逸豪 黄敬英[2] 范勤勤 WANG Yihao;HUANG Jingying;FAN Qinqin(Logistics Research Center,Shanghai Maritime University,Shanghai 201306,China;Postanesthesia Care Unit,Sir Run Run Shaw Hospital,Zhejiang University School of Medicine,Hangzhou 310020,Zhejiang,China)

机构地区：[1]上海海事大学物流研究中心,上海201306 [2]浙江大学医学院附属邵逸夫医院麻醉恢复室,浙江杭州310020

出　　处：《陕西师范大学学报（自然科学版）》2023年第5期25-34,共10页Journal of Shaanxi Normal University：Natural Science Edition

基　　金：上海市浦江人才计划(22PJD030);国家自然科学基金山东联合基金(U2006228)。

摘　　要：特征选择中特征数量和分类精度之间的关系通常可以看作是一个多模态多目标优化问题,但现有大多数多模态多目标进化算法对于高维优化问题的求解存在搜索能力不足的问题。为解决该问题,提出一种基于因果模型和多模态多目标进化算法的两阶段特征选择方法。在该方法中,首先使用因果模型对数据进行特征选择以便降低问题维度;然后使用多模态多目标优化算法搜索具有多模态特性的特征子集。为验证所提算法性能,它被用于术中低体温风险预测模型的特征选择问题。实验结果表明,提出的两阶段特征选择方法不仅融合了2种不同方法的优点,而且能为术中低体温预测提供更多决策支持。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.

关 键 词：因果模型 多模态多目标优化 特征选择 术中低体温 进化计算 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]