The objective of this research was to develop operational travel demand forecasting models consistent with travel choice behavior and with coefficients estimated by use of data at the level of households or individual travelers. It was determined that the three most critical issues in applying such disaggregate models were: (1) whether the resulting models are transferable, (2) how to aggregate the model results for forecasting, and (3) whether the Independence of Irrelevant Alternatives property was an impediment to application of the resulting models. Findings on these critical issues together with a description of the state of the art in application of disaggregate models are given in this report. Under certain conditions and problem requirements (e.g., accuracy and area scale), a disaggregate model developed for one geographic area may be transferred to another area, but some adjustments to the model (e.g., recalibrating mode-specific constants), so that the model predictions match observed aggregate model splits, are necessary in most instances. The authors suggest that aggregation is best handled by segmenting travelers into similar groups on the basis of factors important to explaining travel behavior and its variability. The model would be applied to each segment and aggregated. With regard to the independence of the irrelevant alternative assumption, the authors state that the assumption can be a desirable and reasonable one for homogeneous traveler segments. For these issues, guidance is provided analysts in developing an approach to using disaggregate models for policy analysis.
Samenvatting