Deep neural networks for choice analysis: Extracting complete economic information for interpretation
While deep neural networks (DNNs) have been increasingly applied to choice analysis showing high predictive power, it is unclear to what extent researchers can interpret economic information from DNNs. This paper demonstrates that DNNs can provide economic information as complete as classical discre...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
Elsevier BV
2020
|
| Online Access: | https://hdl.handle.net/1721.1/127230 |
| _version_ | 1811095207650787328 |
|---|---|
| author | Wang, Shenhao Wang, Qingyi Zhao, Jinhua |
| author2 | Massachusetts Institute of Technology. Department of Urban Studies and Planning |
| author_facet | Massachusetts Institute of Technology. Department of Urban Studies and Planning Wang, Shenhao Wang, Qingyi Zhao, Jinhua |
| author_sort | Wang, Shenhao |
| collection | MIT |
| description | While deep neural networks (DNNs) have been increasingly applied to choice analysis showing high predictive power, it is unclear to what extent researchers can interpret economic information from DNNs. This paper demonstrates that DNNs can provide economic information as complete as classical discrete choice models (DCMs). The economic information from DNNs includes choice predictions, choice probabilities, market shares, substitution patterns of alternatives, social welfare, probability derivatives, elasticities, marginal rates of substitution, and heterogeneous values of time. Unlike DCMs, DNNs can automatically learn utility functions and reveal behavioral patterns that are not prespecified by domain experts, particularly when the sample size is large. However, the economic information obtained from DNNs can be unreliable when the sample size is small, because of three challenges associated with the automatic learning capacity: high sensitivity to hyperparameters, model non-identification, and local irregularity. The first challenge is related to the statistical challenge of balancing approximation and estimation errors of DNNs, the second to the optimization challenge of identifying the global optimum in the DNN training, and the third to the robustness challenge of mitigating locally irregular patterns of estimated functions. To demonstrate the strength and challenges, we estimated the DNNs using a stated preference survey from Singapore and a revealed preference data from London, extracted the full list of economic information from the DNNs, and compared them with those from the DCMs. We found that the economic information either aggregated over trainings or population is more reliable than the disaggregate information of the individual observations or trainings, and that larger sample size, hyperparameter searching, model ensemble, and effective regularization can significantly improve the reliability of the economic information extracted from the DNNs. Future studies should investigate the requirement of sample size, better ensemble mechanisms, other regularizations and DNN architectures, better optimization algorithms, and robust DNN training methods to address DNNs three challenges to provide more reliable economic information for DNN-based choice models. |
| first_indexed | 2024-09-23T16:12:30Z |
| format | Article |
| id | mit-1721.1/127230 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T16:12:30Z |
| publishDate | 2020 |
| publisher | Elsevier BV |
| record_format | dspace |
| spelling | mit-1721.1/1272302022-09-29T18:58:05Z Deep neural networks for choice analysis: Extracting complete economic information for interpretation Wang, Shenhao Wang, Qingyi Zhao, Jinhua Massachusetts Institute of Technology. Department of Urban Studies and Planning While deep neural networks (DNNs) have been increasingly applied to choice analysis showing high predictive power, it is unclear to what extent researchers can interpret economic information from DNNs. This paper demonstrates that DNNs can provide economic information as complete as classical discrete choice models (DCMs). The economic information from DNNs includes choice predictions, choice probabilities, market shares, substitution patterns of alternatives, social welfare, probability derivatives, elasticities, marginal rates of substitution, and heterogeneous values of time. Unlike DCMs, DNNs can automatically learn utility functions and reveal behavioral patterns that are not prespecified by domain experts, particularly when the sample size is large. However, the economic information obtained from DNNs can be unreliable when the sample size is small, because of three challenges associated with the automatic learning capacity: high sensitivity to hyperparameters, model non-identification, and local irregularity. The first challenge is related to the statistical challenge of balancing approximation and estimation errors of DNNs, the second to the optimization challenge of identifying the global optimum in the DNN training, and the third to the robustness challenge of mitigating locally irregular patterns of estimated functions. To demonstrate the strength and challenges, we estimated the DNNs using a stated preference survey from Singapore and a revealed preference data from London, extracted the full list of economic information from the DNNs, and compared them with those from the DCMs. We found that the economic information either aggregated over trainings or population is more reliable than the disaggregate information of the individual observations or trainings, and that larger sample size, hyperparameter searching, model ensemble, and effective regularization can significantly improve the reliability of the economic information extracted from the DNNs. Future studies should investigate the requirement of sample size, better ensemble mechanisms, other regularizations and DNN architectures, better optimization algorithms, and robust DNN training methods to address DNNs three challenges to provide more reliable economic information for DNN-based choice models. 2020-09-10T15:34:07Z 2020-09-10T15:34:07Z 2020-09 2020-02 2020-08-31T12:46:48Z Article http://purl.org/eprint/type/JournalArticle 0968-090X https://hdl.handle.net/1721.1/127230 Wang, Shenhao, Qingyi Wang and Jinhua Zhao. “Deep neural networks for choice analysis: Extracting complete economic information for interpretation.” Transportation Research Part C: Emerging Technologies, 118 (September2020): 102701 © 2020 The Author(s) en 10.1016/j.trc.2020.102701 Transportation Research Part C: Emerging Technologies Creative Commons Attribution-NonCommercial-NoDerivs License http://creativecommons.org/licenses/by-nc-nd/4.0/ application/pdf Elsevier BV Other repository |
| spellingShingle | Wang, Shenhao Wang, Qingyi Zhao, Jinhua Deep neural networks for choice analysis: Extracting complete economic information for interpretation |
| title | Deep neural networks for choice analysis: Extracting complete economic information for interpretation |
| title_full | Deep neural networks for choice analysis: Extracting complete economic information for interpretation |
| title_fullStr | Deep neural networks for choice analysis: Extracting complete economic information for interpretation |
| title_full_unstemmed | Deep neural networks for choice analysis: Extracting complete economic information for interpretation |
| title_short | Deep neural networks for choice analysis: Extracting complete economic information for interpretation |
| title_sort | deep neural networks for choice analysis extracting complete economic information for interpretation |
| url | https://hdl.handle.net/1721.1/127230 |
| work_keys_str_mv | AT wangshenhao deepneuralnetworksforchoiceanalysisextractingcompleteeconomicinformationforinterpretation AT wangqingyi deepneuralnetworksforchoiceanalysisextractingcompleteeconomicinformationforinterpretation AT zhaojinhua deepneuralnetworksforchoiceanalysisextractingcompleteeconomicinformationforinterpretation |