Comparative analysis with traditional statistical methods and artificial neural networks of trip generation models according to households characteristics

Trip generation is the first step of transportation planning and trip distribution, stochastic separation, and assignment of traffic are done at the end of the finding of the values. For this reason, with the accurate and thorough calculation of trip generation, which is the basis of planning, it can be maintained that the planning steps to be applied later will be carried out in a healthy way and the investments that are made will be relevant and the costs will be minimum. In the scope of this study, 3 provinces were selected from developed, developing, and non-developed provinces in Turkey and the field of the study was determined. In the determined provinces, trip generation models were generated according to the household characteristics by making household transportation surveys in the determined provinces. The aim of this study is to determine the trip generation of provinces of different categories according to the household characteristics related to the size and development situation of the provinces and to determine the factors affecting trip generation in these provinces. Besides, it is decided which one of the linear, poisson, and negative binomial regression models is more appropriate for trip generation and Artificial Neural Networks model is compared with the most significant regression model. At the end of the analyses, Artificial Neural Network models have shown better performance among three different data sets. As a result, Artificial Neural Networks were proposed as an alternative method in the trip generation of the provinces.