Abstract
Recognizing the limitations of previous travel mode choice models such as random utility models and multi-layer perceptron neural network models, this study develops a framework using a deep neural network with deep learning schemes, to predict travelersâ?? mode choice behavior. Deep neural networks and deep learning are relatively newer models, applied mostly so far to pattern recognition and image/voice processing, and for big data analytics. The authors develop such a choice model with a structure that is appropriate for the travel mode choice problem, and demonstrate the success of the model using an available dataset. The research also develops an important component of the model that takes into account the inherent characteristics of choice models that all individuals have different choice alternatives, an aspect not considered in the neural network models of the past that led to poorer performance. The proposed model is compared against existing mode choice models. The results prove that the new model clearly outperforms the previous mode choice models.