Research On Urban Residents’ Travel Chain Forecasting Based On Multi-Level Nested Logit Model

Currently, the major domestic and international travel demand forecasting model is based on the four-stage method. The tradition Four-step method takes aggregate modeling concept as the basis and the zones as the research targets, it studied the overall characteristics of the zones travel, ignoring the individual travel behavior characteristics. Because of the amazing speed of the economy development, the expansion of the urbanization and the increasing number of residents travel, the Four-stage method of disaggregate model cannot satisfy our planning demand. The disaggregate model studies the individual travel behavior and can truly reflect the individual’s trip choice.As the traditional trip generation, distribution and mode split technologies have some disadvantages, this paper will establish the Multi-level Nested Logit model of travel demand forecasting. This paper will present this trip generation-destination-mode choice with Nested Logit model through selecting variables and computing as well as testing the parameters from the bottom to the top of this NL model to optimize trip demand modeling system and truly reflect this choice process.The main contents of this paper include the following:1 The paper analyzes the defects of the traditional four stage method, and proposes the idea of disaggregate joint selection model. At the same time, the basic theory of the disaggregate model is introduced, and the paper elaborates MNL model and NL model in detail.2 According to the survey data, the residents travel characteristics are analyzed, it analyzes the influence of different factors on trip mode and trip times from personal attributes, family property, travel characteristics etc.,3 On the basis of summing up the basic theoretical knowledge, combined with the survey data of residents’ travel, the model system of the resident travel demand forecasting is constructed, and a three-level NL model framework based on MNL model is formed. The system includes a travel mode choice model, destination choice model and trip generation model, it describes the probability formula and the formula of the utility function, and achieves linkage and feedback of each model through Logsum variable.4 The basic principle, influencing factors (SPSS), the utility function settings, coefficient calibration and tests of each sub-model are explained in detail and description, and the results of parameter calibration are also explained in each stage. Finally, the rationality of the whole model system is tested by calculating the sharing rate of each trip mode.