2010 Volume 8 Pages 603-614
Transportation demand models estimate travel demands due to daily activities of persons. The gravity model is the representative approach of trip distribution drawn between the spatial interaction of trip making and travel costs. However, the standard gravity model shows the modeling limitation due to distance-based distribution in cases of polycentric city structure. Many transportation researchers insist that the gravity model should be improved by concerning new variables of areal characteristics or interzonal relative attractiveness. This research develops an estimation procedure for interzonal relative attractiveness as a new variable for explaining zonal spatial properties and interzonal spatial association using 1996 and 2002 interzonal commuting trips of the SM city. The research is extended to develop an improved trip distribution model containing the counterpropagation neural network with three hidden layers. The applicability of the improved model is evaluated by four criteria including β, E-norm, RMSE, and Theil value.