Research of multilevel models for demand forecast of urban rail transit

For the huge-investment project like rail transit in city, the forecast passenger demand is very important to its planning and feasibility studying. Traditional forecasting methods or models can not fully use all the survey data in passenger demand forecast, and some information is wasted. They have another deficiency that receives a low accuracy value in demand forecasting for incomplete factors are taken in count. Try to solve these deficiency, this paper proposes a model, multi-level model, which considers multilevel utility and the different characteristic of individual that from a define aggregate of total group. It n will separate variation of zones from residual of traditional Logit model, to improve forecast accuracy, and deepened the understanding of the system architecture level. The 2-level Logit model considering utility in this paper is a more suitable model than traditional statistic models. It has a higher accuracy, and can quotas a higher level model as well. The sample shows that the 2-level Logit model can well simulate and precast the transit demand in a convenient way to transit planning and designing.