基于CART决策树的乘客疏散行为决策影响因素研究

Research on Influencing Factors of Passenger Evacuation Behavior Decision Making Based on CART Decision Tree

摘要: 目的：为保障乘客安全和地铁安全运营，确保地铁突发火灾事故下乘客的有序疏散，对乘客的疏散行为影响因素进行分析。方法：首先，依托行为决策理论与识别乘客疏散行为影响因素；其次，利用机器学习CART决策树对不同类别乘客的异质化疏散行为决策进行分类，并对影响因素的相对重要性进行排序。随后，根据最显著影响因素拥挤疏散路口的滞留等待时长，构建Anylogic仿真模型，模拟地铁突发火灾事故下的乘客疏散行为。结果及结论：模拟结果验证了优化策略的有效性，表明在站厅拥堵处引入疏散引导员，能够平衡各出口疏散乘客，显著缩短疏散时长，抑制乘客负面情绪产生，有效缩短疏散时间。

Abstract: Purpose: To protect passenger safety and the safe operation of the subway, and to ensure the orderly evacuation of passengers under the subway fire accident, we analyze the factors influencing the evacuation behavior of passengers. Methods: first, relying on behavioral decision-making theory to identify passengers’ evacuation behavioral influencing factors; second, using machine learning CART decision tree to classify different categories of passengers’ heterogeneous evacuation behavioral decisions and rank the relative importance of influencing factors. Subsequently, an Anylogic simulation model is constructed to simulate the passenger evacuation behavior under a subway fire accident based on the length of detention waiting at crowded evacuation intersections, which is the most significant influencing factor. Results and Conclusion: The simulation results verify the effectiveness of the optimization strategy and show that the introduction of evacuation guides at congested station halls can balance the evacuation of passengers at all exits, significantly shorten the evacuation time, inhibit the generation of negative emotions among passengers, and effectively shorten the evacuation time.