An Analysis On Uncertainty Propagation In The Four-step Travel Demand Forecasting Model

The influence of uncertainty in transportation demand forecasts should be studied in order to make correct decision since the transportation system is a complicated system comprising of much uncertainty. Uncertainty in traffic demand forecast models caused more attention by the foreign planners. Research shows that demand uncertainty is the essential attribute of the prediction model, and its research is very few internally. This study provides a framework for analyzing and quantifying the uncertainty involved in travel demand forecasting models.This article first reviewed the commonly used four-step travel demand forecasting models and model the impact of factors, the selection of model parameters and so on, then for specific four-step travel demand forecasting models, analysised the uncertainty in-depth, defined the sources of uncertainty in the four-step model, and classified the uncertainty of prediction models into two categories mainly: input uncertainty and model uncertainty.For four-step model uncertainties features, selected the two dimensions Monte Carlo simulation analysis method. This method is able to quantitatively distinguish the general uncertainty caused by the input variable uncertainty and the model parameters uncertainty, in order to help decision makers choose the factors that have significant impact on the model results , targeted to improve the situation.Used a simple network as study case, for the four-step model uncertainty problem, consider the number of people, mode choice, distribution impendence coefficient as the risk variable, quantitative analysised the output change of each stage. After analysised the example network, found that compared with the model input, the model parameters contribute major model output uncertainty. When predict the future situation and decision making should pay more attention to the four-step model's parameters. The example also quantitative analysised the uncertainty propagation of the four-step model. The conclusion was that the uncertainty of the first three steps of a four-step model is cumulative occurrence, but if the final assignment stage uses equilibrium distribution method would reduce the uncertainty of the link flow.