Mixed Logit模型随机参数分布形式选择探讨

Study on the Choice of Random Parameter Distribution Function of Mixed Logit Model

MixedLogit模型参数的分布形式的选择对模型物理意义的阐释和模型的实际应用均有十分重要的影响。选取正态分布、对数正态分布和均匀分布3种代表性分布函数进行分析,以Daganzo1979年应用Logit模型进行交通方式划分模型标定所采用的原始数据进行试验研究,并将试验结果与Logit模型结果进行对比。研究结果表明:MixedLogit模型对原始数据的拟合性更好;3种分布函数分别适应于不同特点的参数;MixedLogit具有更为丰富的物理意义,对个人选择交通方式的偏好信息表达更为清晰。

The choice of the random parameter distribution function of Mixed Logit Model is key to the use of the model in practice. Normal distribution, log normal distribution and uniform distribution are selected and analyzed as the distribution function of the random parameter by use of the data from Daganzo who made use of the data to analyze the mode split employing Logit Model. The results are compared between Logit Model and Mixed Logit Model, and also among the three different distribution functions. The results shows that Mixed Logit Model fits the data better than Logit Model, and the three different distribution in study is adapt to different kind of parameter. In addition, mixed logit model is better in expressing the personal preference for travel mode choice.