A Logit Model with Non-Linear Utility Functions

The Exemplum of Utility Functions consisting of Neural Networks

Abstract

In most of logit models, linear utility functions have frequently been used. We propose a utility function with neural network, and make a logit model with it. We apply the model to the giveway behavior at the merging point on the road, and examine the validity of the model. As a result, it is founded that the logit model with neural network utility function describes the giveway behavior more elaborately than the conventional logit model and its goodness of fit is also better. Thus, we confirm that the logit model with neural network utility function is valid.