突发事件下人群疏散交通方式选择仿真与分析  

Simulation and analysis on traffic mode choice of crowd evacuation under emergency

作　　者：张庆宇 李韧 茹含 丁庆涛 李孟晖 ZHANG Qingyu;LI Ren;RU Han;DING Qingtao;LI Menghui(College of Water Resources and Architectural Engineering,Northwest A&F University,Yangling Shaanxi 712100,China;China Harbour Engineering Company Ltd.,Beijing 100027,China)

机构地区：[1]西北农林科技大学水利与建筑工程学院,陕西杨凌712100 [2]中国港湾工程有限责任公司,北京100027

出　　处：《中国安全生产科学技术》2023年第7期162-168,共7页Journal of Safety Science and Technology

基　　金：西北农林科技大学大学生科创基金项目(S202210712620);特殊地区公路工程教育部重点实验室(长安大学)开放基金项目(300102211502);西北农林科技大学中央高校学科基金项目(2452020238)。

摘　　要：为制定合理的疏散资源分配方案,建立基于不同社会背景集群的人群疏散交通方式选择模型。通过分析突发事件下人群疏散交通方式选择的影响因素,设计陈述性偏好(SP)调查问卷,得到样本在突发事件下的选择偏好基础数据;采用因子分析法对数据进行降维处理,采用聚类算法最大程度消除个体异质性的影响;基于多项Logit回归分析和SVM模型预测,分别建立基于不同社会背景集群的突发事件人群疏散交通方式选择模型。研究结果表明:不同集群之间的疏散方式选择模型存在差异性,这是由数据集的异质性造成的,本文建立的选择模型可尽量避免异质性对模型预测结果的影响;多项Logit模型总预测准确率达76.32%,比传统非聚类模型提升15.54%;SVM模型总预测准确率达83.85%,比非聚类模型提升8.95%。研究结果可为管理者制定疏散资源分配方案提供依据。In order to formulate the allocation scheme of evacuation resources,a traffic mode choice model of crowd evacuation for different social background clusters was established.By analyzing the influencing factors for the traffic mode choice model of crowd evacuation under emergency,the stated preference(SP)questionnaire was designed to obtain the basic data of sample choice preference under emergency.The factor analysis method and clustering algorithm were used to assign the survey objects to the homogeneous clusters based on the crowd clustering of different social backgrounds,so as to eliminate the influence of individual heterogeneity on the accuracy of prediction results to the greatest extent.Through the multinomial Logit regression analysis and SVM model prediction,the traffic mode choice model of crowd evacuation under the corresponding emergency of each cluster was established respectively.The results show that there are differences in the evacuation mode choice mode between different clusters,which is caused by the heterogeneity of data set.The selection model established in this paper can avoid the influence of heterogeneity on the prediction results of model as possible.The total prediction accuracy of the multinomial Logit model reaches 76.32%,which is 15.54%higher than that of the traditional non-clustering model.The total prediction accuracy of SVM model reaches 83.85%,which is 8.95%higher than that of non-clustering model.The results can provide basis for managers to formulate the allocation scheme of evacuation resources.

关 键 词：突发事件 交通方式选择 SP调查 聚类分析 LOGIT模型 支持向量机 人群疏散 

分 类 号：X951[环境科学与工程—安全科学]