A structural and comparative analysis for various trip production models, case study of Qazvin educational trips

Trip generation is the first stage of the conventional four-step travel forecasting framework that estimates the number of trips to and from a traffic analysis zone. Using linear regression model is common in this step and generates an acceptable level of performance from the perspective of transport planning, however this model does not incorporate traveler behavior, integer and non-negative nature of trips. To overcome these limitation, several models have been suggested: censored model such as Tobit for deleting negative values; count data models such as negative binomial and Poisson for deleting continuous and negative values; and discrete choice models such as ordered logit and probit for incorporating traveler behavior and preventing continuous and negative values. Given the importance of trip generation stage and lack of sufficient and quantitive attention to various trip production models, this paper develops alternative trip production models. The purpose of this paper is a structural analysis for various trip production models and comparison of their performance in prediction. Four representative models (regression, Tobit, Poisson and ordered logit) are applied to the educational trips in Qazvin city. The modeling unit employed in this study is the household. Sample is included econometric- social attributes 4734 houshols. 85% of the data used for estimation and the rest to validation. The models are assessed by how closely they are able to replicate trips ,made by each household in the estimation and validation dataset. in order to compare the performance of models in prediction, each of the models is developed on estimation dataset, and the models are used to predict the trips made by each household in validation and estimation dataset. Measures assessing how well the predicted number of trips made daily by each houshold by each of the models compared to the observed number of trip made by the houshold are evaluated and compared. The four measure for assessing performance are the mean absolute error, regression of the predicted number of household- trips against the observed number of household trip in term of goodness of fit and coefficient of determination, and compare plot of observed and predicted aggregate trip shares. In order to modeling is used stata software. The result show that, In every four models, number of school students, number of university students, and household car ownership have been statistically significant. The performance of each of the models are different in term of various measures (mean absolute error, regression of the predicted number of household- trips against the observed number of household trip in term of goodness of fit and coefficient of determination, and compare plot of observed and predicted aggregate trip shares).From mean absolute error perspective, ordered Logit and linear regression models have the best performance, but from goodness of fit regression of the predicted number of household- trips against the observed number of household trip, Tobit models have the best performance. Ordered Logit models have the best performance in terms of coefficient of determination of the predicted number of household- trips against the observed number of household trip and comparision of predicted share of every trip rate level with observed share. The performance of each of the models are similar in prediction of validation and estimation dataset.