Comparison analysis of travel demand forecasting model

Abstract

Reasonable transportation planning is a fundamental solution of traffic jams in big cities and the core is travel demand forecast. Reasonable travel demand forecast will provide scientific evidence for decision of transportatio n planning. Thus, it seems very urgent to do some research on travel demand forecast in varies countries. In this paper, we first summarize most of research work on the travel demand forecasting model in domestic and overseas. Then, discuss the future work in this field. Introduction of Aggregated Model In the early 60's of last century, 《Chicago Area Transportation Study》 contributed to the formulation and development of transportation planning theory and method. This law pointed out that urban transportation planning should be based on traffic survey to obtain the financial subsidies of highway construction from federal governments. The earliest travel demand forecasting model was simple and aggregated model, contains several parameters, and using aggregated data to estimate them. The representation of aggregated model is "four stage model", which is formulated in the 70's of last century. The model contains trip generation, distribution, mode split and traffic assignment. After that, many scholars have inducted the theory of other disciplines to made corresponding improvement to four stage model. Considering to the data collection, all those improved models still takes traffic district as a basic analysis unit(1)。Thus, the defects of traditional aggregated model are as follows: 1) lack of economic theory base, social, Psychology and behavior analysis. 2) lack of consistency between each stage, no feedback mechanism. 3) lack of statistical method 4) lack of evaluation mechanism between transportation management and demand 5) forecasting ability is relatively weak. Considering on the above defects of four stage model, scholars have studied and developed more complex model which is disaggregated and more fit with real demand. This type of model is based on the theory of probability, economics, sociology and behavioral science, and accurately describe the selection process of person and household travel(2). The representation of disaggregated model is Logit model which is deducted from Luce in 1959. After that, McFadden who is from MIT has discussed the logit model and it's features, then formed the theoretical system, put them into the practical stage in 1974(3). In the last 70's of last century, the basic idea and methods of disaggregated model were applied to the actual forecast and be improved in 80's. Since then, along with the development of computer technology and simulation algorithm, there have been some more complicated and abundant disaggregated models, such as Probit model, Mixed Logit model, OGEV model(Ordered generalized extreme value), PCL model(Paired combinatorial logit), MNP model (Multinomial Probit)and CNL model(Cross-nested logit)and so on(4).