Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Workplace Location Choice Model based on Deep Neural Network

Tanay Rastogi, Anders Karlström

TL;DR

This study addresses workplace location choice by comparing a traditional 2-level Nested Logit DCM with two DNN-based models (DNN-Car and DNN-All) using Stockholm travel survey and SAMS data. The DCM employs spare time accessibility and occupation-specific terms to estimate choice probabilities via $P_{n,j|i} = \frac{e^{v_{n,ij}}}{\sum_{j} e^{v_{n,ij}}}$, while the DNNs use a zone-block architecture with an ASC layer and a softmax output to produce $\Pr(j|i)$, trained with Adam optimization. Results show the DNNs achieve lower training and validation log-likelihoods, with DNN-All outperforming DNN-Car, indicating richer individual data improves predictions; however, the DCM better matches short-distance patterns and some distance-related attributes, as evidenced by KS tests. Overall, the findings highlight the potential of data-driven DNNs as robust alternatives to traditional DCMs in workplace location analysis, while underscoring the value of model choice based on the specific planning or policy application and the importance of hybrid approaches for future ILUTMs.

Abstract

Discrete choice models (DCMs) have long been used to analyze workplace location decisions, but they face challenges in accurately mirroring individual decision-making processes. This paper presents a deep neural network (DNN) method for modeling workplace location choices, which aims to better understand complex decision patterns and provides better results than traditional discrete choice models (DCMs). The study demonstrates that DNNs show significant potential as a robust alternative to DCMs in this domain. While both models effectively replicate the impact of job opportunities on workplace location choices, the DNN outperforms the DCM in certain aspects. However, the DCM better aligns with data when assessing the influence of individual attributes on workplace distance. Notably, DCMs excel at shorter distances, while DNNs perform comparably to both data and DCMs for longer distances. These findings underscore the importance of selecting the appropriate model based on specific application requirements in workplace location choice analysis.

Workplace Location Choice Model based on Deep Neural Network

TL;DR

This study addresses workplace location choice by comparing a traditional 2-level Nested Logit DCM with two DNN-based models (DNN-Car and DNN-All) using Stockholm travel survey and SAMS data. The DCM employs spare time accessibility and occupation-specific terms to estimate choice probabilities via , while the DNNs use a zone-block architecture with an ASC layer and a softmax output to produce , trained with Adam optimization. Results show the DNNs achieve lower training and validation log-likelihoods, with DNN-All outperforming DNN-Car, indicating richer individual data improves predictions; however, the DCM better matches short-distance patterns and some distance-related attributes, as evidenced by KS tests. Overall, the findings highlight the potential of data-driven DNNs as robust alternatives to traditional DCMs in workplace location analysis, while underscoring the value of model choice based on the specific planning or policy application and the importance of hybrid approaches for future ILUTMs.

Abstract

Discrete choice models (DCMs) have long been used to analyze workplace location decisions, but they face challenges in accurately mirroring individual decision-making processes. This paper presents a deep neural network (DNN) method for modeling workplace location choices, which aims to better understand complex decision patterns and provides better results than traditional discrete choice models (DCMs). The study demonstrates that DNNs show significant potential as a robust alternative to DCMs in this domain. While both models effectively replicate the impact of job opportunities on workplace location choices, the DNN outperforms the DCM in certain aspects. However, the DCM better aligns with data when assessing the influence of individual attributes on workplace distance. Notably, DCMs excel at shorter distances, while DNNs perform comparably to both data and DCMs for longer distances. These findings underscore the importance of selecting the appropriate model based on specific application requirements in workplace location choice analysis.

Paper Structure

This paper contains 15 sections, 8 equations, 5 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Workplace Choice Models
  3. Discrete Choice Model (DCM)
  4. Deep Neural Network (DNN)
  5. Data and Estimation Results
  6. Data
  7. Training Results
  8. DCM
  9. DNN
  10. Model Comparison
  11. Workplace Attributes
  12. Individual Attributes
  13. Conclusion
  14. Future Works
  15. Acknowledgment

Figures (5)

  • Figure 1: 2-level NL model for workplace location choice model proposed by Naqavi2023MobilityStockholm.
  • Figure 2: Structure of the proposed DNN for workplace location choice model.
  • Figure 3: Total number of workplaces in each of 1375 zones in Stockholm.
  • Figure 4: Probability distribution of distance of chosen work from given home.
  • Figure 5: Probability distribution of distance of chosen work from given home based on the individual attributes.