Deep Neural Network Design for Modeling Individual-Level Travel Mode Choice Behavior

Individual-level modeling is an essential requirement for effective deployment of smart urban mobility applications. Mode choice behavior is also a core feature in transportation planning models, which are used for analyzing future policies and sustainable plans such as greenhouse gas emissions reduction plans. Specifically, an agent-based model requires an individual level choice behavior, mode choice being one such example. However, traditional utility-based discrete choice models, such as logit models, are limited to aggregated behavior analysis. This paper develops a model employing a deep neural network structure that is applicable to the travel mode choice problem. This paper uses deep learning algorithms to highlight an individual-level mode choice behavior model, which leads us to take into account the inherent characteristics of choice models that all individuals have different choice options, an aspect not considered in the neural network models of the past that have led to poorer performance. Comparative analysis with existing behavior models indicates that the proposed model outperforms traditional discrete choice models in terms of prediction accuracy for both individual and aggregated behavior.

[1]  Andrea Papola,et al.  Random utility models with implicit availability/perception of choice alternatives for the simulation of travel demand , 2001 .

[2]  Taghi M. Khoshgoftaar,et al.  Deep learning applications and challenges in big data analytics , 2015, Journal of Big Data.

[3]  Michel Bierlaire,et al.  A theoretical analysis of the cross-nested logit model , 2006, Ann. Oper. Res..

[4]  Peng Jing,et al.  Travel Mode and Travel Route Choice Behavior Based on Random Regret Minimization: A Systematic Review , 2018 .

[5]  Yee Whye Teh,et al.  A Fast Learning Algorithm for Deep Belief Nets , 2006, Neural Computation.

[6]  Francisco C. Pereira,et al.  A Neural-embedded Choice Model: TasteNet-MNL Modeling Taste Heterogeneity with Flexibility and Interpretability , 2020, ArXiv.

[7]  David A. Hensher,et al.  A comparison of the predictive potential of artificial neural networks and nested logit models for commuter mode choice , 1997 .

[8]  Eric J. Miller,et al.  Nested Logit Models and Artificial Neural Networks for Predicting Household Automobile Choices: Comparison of Performance , 2002 .

[9]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[10]  Peter Vovsha,et al.  Application of Cross-Nested Logit Model to Mode Choice in Tel Aviv, Israel, Metropolitan Area , 1997 .

[11]  K. Train A Structured Logit Model of Auto Ownership and Mode Choice , 1980 .

[12]  Yunlong Zhang,et al.  Travel Mode Choice Modeling with Support Vector Machines , 2008 .

[13]  T F Golob,et al.  THE EFFECTIVENESS OF RIDESHARING INCENTIVES: DISCRETE-CHOICE MODELS OF COMMUTING IN SOUTHERN CALIFORNIA. REVISED EDITION , 1991 .

[14]  Gaël Varoquaux,et al.  Scikit-learn: Machine Learning in Python , 2011, J. Mach. Learn. Res..

[15]  Giulio Erberto Cantarella,et al.  Multilayer Feedforward Networks for Transportation Mode Choice Analysis: An Analysis and a Comparison with Random Utility Models , 2005 .

[16]  J. S. Long,et al.  Testing for IIA in the Multinomial Logit Model , 2007 .

[17]  Amer Shalaby,et al.  INVESTIGATING THE ROLE OF RELATIVE LEVEL-OF-SERVICE CHARACTERISTICS IN EXPLAINING MODE SPLIT FOR THE WORK TRIP , 1998 .

[18]  Hichem Omrani,et al.  Predicting Travel Mode of Individuals by Machine Learning , 2015 .

[19]  K. Small A Discrete Choice Model for Ordered Alternatives , 1987 .

[20]  Anjali Awasthi,et al.  Prediction of Individual Travel Mode with Evidential Neural Network Model , 2013 .

[21]  M. C. M. de Carvalho,et al.  Forecasting travel demand: a comparison of logit and artificial neural network methods , 1998, J. Oper. Res. Soc..

[22]  Luís A. Alexandre,et al.  Data classification with multilayer perceptrons using a generalized error function , 2008, Neural Networks.