An Improvement of the Binary Logit Model for Trip Generation Forecasting

This paper developed a hypothesis to overcome the disadvantage of the binary logit model used in resident trip generation. It proposed that random parameters beta' obeyed some distribution firstly. Combined the characters of trip generation and observable variables, it then put forward that beta' followed log normal distribution. The hypothesis was validated by demarcating the parameters in SAS software and forecasting trip generation volumes with two different models, both of which were based on PUMS investigation data in USA. The former result shows that all the parameters' T values are out of [-1,1] interval and proves the assumption is correct. The latter one shows that the improved model's forecasting precision is much higher than the previous one's, and reveals that the application of the improved binary logit model in resident trip generation has notable effects.