交通方式划分离散选择模型的比较研究

摘要

Five typical discrete choice models were introduced, which were MNL, NL, HEV, MNP and MMNL.The model characteristics and their features were discussed and compared. Then these models were used to study the mode split in a case, and to calibrate the model parameters. The result indicated that each model would have different representation in the same case. It was also found that NL model could solve the similarity in the alternative heterogeneity. Results from this analysis suggested that the error of MMNL was smallest,its prediction reliability was highest. MMNL model was a better model than the other four models in solving the heterogeneity and similarity problems and it could improve the model explanation power significantly,and was the most effective method to predict mode choice. The research is very important for transport mode split forecast, which indicates that the conclusion derived from only one model in the existing researches may be not reliable.It should be chose cautiously according to the characteristics of the models.%本文章介绍了MNL、NL、HEV、MNP、MMNL等五种离散选择模型，重点分析了它们各自的理论基础与特点，并应用于交通方式划分案例，对模型参数进行标定。结果表明，NL模型能够较好解决存在于替选方案的相似性问题；MMNL预测误差最小，可靠性最高，较NL、HEV、MNP模型更能解决交通方式选择行为中存在的相似性、异质性问题，是一种对交通方式选择行为预测的有效方法。该研究对交通方式分担率预测问题具有重要的指导意义，因为通过简单选择某个离散选择模型开展研究得出的结论并不一定可靠，应根据模型特点谨慎选择、谨慎应用。