Overcrowding in metro systems can exert a negative impact on the travel experience of the passengers and on the system’s capacity. This can affect passengers in different ways, worsening their perception of safety, altering their behaviour, decreasing their satisfaction, and leading to higher travel-time sensitivities. One of the causes of overcrowding can be attributed to some metro cars being selected more often than others, and thus, the distribution of passengers inside the train not being homogeneous. This study adopts a disaggregated approach based on the observed behaviour to understand why some specific metro cars are preferred. The characteristics of passengers and metro cars that generate heterogeneity in the load profiles are identified and the motives behind passenger choices are understood by formulating a hybrid latent-class choice model. The behavioural models consider sociodemographic characteristics, level-of-service information, and attitudinal latent variables. Once a metro car choice model was obtained, an operational micro-simulator of the metro was programmed to evaluate how passengers' choices would be affected by altering variables related to the design of the stations, assessing it based on the current situation. Subsequently, interventions to station layouts are made based on the forecast of the behavioural models, to affect passenger choices and produce more homogeneous passenger distributions in the train. The intervention presented in this study reduced 46.49 h of total travel time during the morning peak hours; the impact was greater when considering individual perceptions. Incentives can be developed for optimising the distribution of passengers in metro cars by understanding passenger choice using a behavioural approach, and therefore, they can improve the passenger experience.

地铁系统过度拥挤会对乘客的出行体验和系统容量产生负面影响。这可能会以不同的方式影响乘客，恶化他们的安全感，改变他们的行为，降低他们的满意度，并导致更高的旅行时间敏感性。过度拥挤的原因之一可以归因于某些地铁车厢的选择频率高于其他车厢，因此列车内乘客的分布不均匀。本研究采用基于观察到的行为的分类方法来了解为什么某些特定的地铁车厢受到青睐。通过制定混合潜在类别选择模型，可以识别在负载曲线中产生异质性的乘客和地铁车辆的特征，并了解乘客选择背后的动机。行为模型考虑社会人口特征、服务水平信息和态度潜在变量。一旦获得地铁车辆选择模型，就可以对地铁的运行微型模拟器进行编程，以评估改变与车站设计相关的变量将如何影响乘客的选择，并根据当前情况进行评估。随后，根据行为模型的预测对车站布局进行干预，以影响乘客的选择并在列车中产生更均匀的乘客分布。本研究中提出的干预措施减少了早高峰时段的总出行时间 46.49 小时；当考虑个人看法时，影响更大。通过使用行为方法了解乘客的选择，可以制定激励措施来优化地铁车厢内的乘客分布，从而改善乘客体验。