Estimation of percentage of pass-by trips generated by a shopping center using artificial neural networks

Pass-by trips are trips made as intermediate stops on the way from an origi n to a primary trip destination. Accurate estimates of the percentage of pa ss-by trips generated by a land use are extremely important for both planne rs and developers. The traditional method of pass-by trip estimation is reg ression modeling with the help of the U.S. Institute of Transportation Engi neers (ITE) Trip Generation manual. This paper also uses data from the Trip Generation manual, and focuses on an alternative methodology based on Arti ficial Neural Networks (ANNs). Use is made of backpropogation, a popular AN N paradigm, and five different architectures of backpropogations are develo ped, tested and compared against three different regression models - linear , log-log and log-linear forms, respectively. The results from the regressi on and ANN-based models are compared in terms of the Root Mean Square of Er rors (RMSE) of predicted values. It is found that the worst ANN prediction RMSE is lower than the best regression prediction RMSE. ANN-based models ha ve the capability of representing the relationship between the percentage o f pass-by trips and the independent variables more accurately than regressi on analysis at no additional monetary costs.