Residents' travel mode choice model

Abstract

Taking urban residents’ travel mode choice behavior as the research object, firstly, this paper resolved the decision mechanism and main influence factors of residents’ travel mode choice. Then, the differences in modeling principle were analised between support vector machine and traditional methods such as neural network, using statistical learning theory. next, a travel mode choice model based on directed acyclic graph and support vector machine (daGsVM) was established. last, comparison on prediction accuracy has been conducted between models based on daG-sVM and neural network using the same set of data. The results show that the daG-sVM model which based on structural risk minimization effectively controls both the empirical risk and confidence range. Its prediction accuracy is higher than that of neural network for almost 9%, which provides a new method for the prediction of residents’ travel mode choice.