A Non-Linear Analysis of Discrete Choice Behavior by the Logit Model (original) (raw)

The logit model based on random utility theory has often been used for discrete choice behavior analysis. In the conventional logit model, it is assumed that variables are independent of each other and that their relationship is linear. In general, the relationship or behavior of the variables is non-linear, and the assumptions of the logit model are not always proper. Therefore, incorporating neural networks, which are suitable to make an analysis non-linearly, to the logit model is very useful. In this study, we propose the logit model using non-linear utility functions with neural network. Then, we analyze a discrete choice behavior in traffic phenomenon by the model, and examine the validity of the model.