New Travel Demand Models with Back-Propagation Network

This paper explores the application of back-propagation network (BPN) to travel demand analysis. Models are developed to simulate travelers\´ inner-city behaviors and all of them adopt BPNs as main paradigm, for its virtue in non-linear analysis and prediction. Compared to the past researches, which were generally based on aggregate data, the models here are more comprehensive and developed based on disaggregate survey data. At first, three categories of models using BPNs are established to respectively realize trip generation, OD estimation and mode choice analysis-the first three steps in classical "four-step" models for travel demand forecasting. Furthermore, the integrated models are researched in two ways. One method is to use a simple combination of the former separate BPN models, and the other is to create a multilayer back-propagation network (MLBPN). Results show that BPN can be a feasible tool for travel demand analysis.