Conventional forecasting methods for estimating the demand for travel between a pair of zones involves using both trip end and trip distribution models together. This has disadvantages, partly because of the difficulty with estimating intrazonal costs, and partly because the method being interactive, it requires much computational effort and the results are not necessarily easy to understand. The simpler process of direct demand modelling estimates trips as a direct function of zonal economic and social characteristics. This avoids the estimation of trip-end models, avoids the interzonal problem and does not involve the complexities of constraining to given trip-end totals. These relative merits of the direct demand method are investigated here, using data from the Nicosia Transportation Study. Methods for comparing trip models have required the use of statistical techniques hitherto not applied in transport. These include cross-validation, which has been employed with both methods to explore issues of robustness, predictive ability and influential data. Results have highlighted areas where the direct demand model appears to have advantages over the gravity model.
Abstract