Among the discrete choice contexts discussed in the transportation literature, route choice is particularly challenging, for which several model structures with different assumptions were developed. Recently, a group of closed-form multinomial models with asymmetric choice probability functions have been proposed to resolve the class imbalance problem in mode choice. However, these asymmetric models can hardly be estimated when the choice sets vary with observations. This paper fills the gap between the asymmetric models and the existing literature, and also adapt the models to route choice by proposing three approaches to further parameterize the shape parameter.

This paper also tests the asymmetric models by two independent GPS datasets: taxi data collected in Guangzhou as well as bicycle data collected in Tel Aviv. Comparative analysis is conducted between taxi drivers and cyclists on their route choice behavior. The results indicate the existence of class imbalance in both cases. Moreover, the asymmetric models have case-dependent performance, and different parameterizations correspond to different interpretations of the asymmetry in route choice.

在交通文献中讨论的离散选择环境中，路线选择尤其具有挑战性，为此开发了几个具有不同假设的模型结构。最近，提出了一组具有非对称选择概率函数的闭式多项式模型来解决模式选择中的类不平衡问题。However, these asymmetric models can hardly be estimated when the choice sets vary with observations. 本文填补了非对称模型与现有文献之间的空白，并通过提出三种进一步参数化形状参数的方法使模型适应路径选择。

本文还通过两个独立的 GPS 数据集测试了非对称模型：在广州收集的出租车数据以及在特拉维夫收集的自行车数据。对出租车司机和骑车人的路径选择行为进行了比较分析。结果表明两种情况下都存在类别不平衡。此外，非对称模型具有依赖于案例的性能，不同的参数化对应于路径选择中对不对称性的不同解释。