Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Explainable Hierarchical Urban Representation Learning for Commuting Flow Prediction

Mingfei Cai, Yanbo Pang, Yoshihide Sekimoto

TL;DR

A heterogeneous graph-based model is developed to generate meaningful region embeddings at multiple spatial resolutions for predicting different types of inter-level OD flows and outperforms existing models in terms of a uniform urban structure.

Abstract

Commuting flow prediction is an essential task for municipal operations in the real world. Previous studies have revealed that it is feasible to estimate the commuting origin-destination (OD) demand within a city using multiple auxiliary data. However, most existing methods are not suitable to deal with a similar task at a large scale, namely within a prefecture or the whole nation, owing to the increased number of geographical units that need to be maintained. In addition, region representation learning is a universal approach for gaining urban knowledge for diverse metropolitan downstream tasks. Although many researchers have developed comprehensive frameworks to describe urban units from multi-source data, they have not clarified the relationship between the selected geographical elements. Furthermore, metropolitan areas naturally preserve ranked structures, like cities and their inclusive districts, which makes elucidating relations between cross-level urban units necessary. Therefore, we develop a heterogeneous graph-based model to generate meaningful region embeddings at multiple spatial resolutions for predicting different types of inter-level OD flows. To demonstrate the effectiveness of the proposed method, extensive experiments were conducted using real-world aggregated mobile phone datasets collected from Shizuoka Prefecture, Japan. The results indicate that our proposed model outperforms existing models in terms of a uniform urban structure. We extend the understanding of predicted results using reasonable explanations to enhance the credibility of the model.

Explainable Hierarchical Urban Representation Learning for Commuting Flow Prediction

TL;DR

A heterogeneous graph-based model is developed to generate meaningful region embeddings at multiple spatial resolutions for predicting different types of inter-level OD flows and outperforms existing models in terms of a uniform urban structure.

Abstract

Commuting flow prediction is an essential task for municipal operations in the real world. Previous studies have revealed that it is feasible to estimate the commuting origin-destination (OD) demand within a city using multiple auxiliary data. However, most existing methods are not suitable to deal with a similar task at a large scale, namely within a prefecture or the whole nation, owing to the increased number of geographical units that need to be maintained. In addition, region representation learning is a universal approach for gaining urban knowledge for diverse metropolitan downstream tasks. Although many researchers have developed comprehensive frameworks to describe urban units from multi-source data, they have not clarified the relationship between the selected geographical elements. Furthermore, metropolitan areas naturally preserve ranked structures, like cities and their inclusive districts, which makes elucidating relations between cross-level urban units necessary. Therefore, we develop a heterogeneous graph-based model to generate meaningful region embeddings at multiple spatial resolutions for predicting different types of inter-level OD flows. To demonstrate the effectiveness of the proposed method, extensive experiments were conducted using real-world aggregated mobile phone datasets collected from Shizuoka Prefecture, Japan. The results indicate that our proposed model outperforms existing models in terms of a uniform urban structure. We extend the understanding of predicted results using reasonable explanations to enhance the credibility of the model.
Paper Structure (28 sections, 9 equations, 6 figures, 6 tables)

This paper contains 28 sections, 9 equations, 6 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Commuting Flow Prediction
  4. Multilevel Regional Representation Learning
  5. Explainable Urban Computing
  6. Preliminaries
  7. Methodology
  8. Hierarchical Urban Graph Construction
  9. Hierarchical Urban Network
  10. Regional Summary Analysis
  11. Experiments
  12. Experimental Setup
  13. Dataset
  14. Commuting Flows
  15. Grid Indicators
  16. ...and 13 more sections

Figures (6)

  • Figure 1: Different types of hierarchical commuting flows. Subfigure \ref{['fig:m2m']} indicates trip flows between grids and grids, while subfigure \ref{['fig:c2m']} shows people flows between cities at higher levels and grids at lower levels.
  • Figure 2: The overview of Hierarchical Urban Network model. In general, the framework includes the pre-message process for initialization of embeddings using urban indicators, urban relation-aware graph convolution and the post-message process for multilevel flow reconstruction. We take a typical urban subgraph for the instance to introduce our proposed message passing method. There are three urban units and two relations, that is, grid B locates at city A and grid C has flows to city A.
  • Figure 3: Feature importance analysis. For both city and grid embeddings, road network had a more significant impact than the other categories. The distribution of POIs was essential to calculate city representations, while the influence of facilities could be neglected for grid embeddings.
  • Figure 4: Grid attributions for city embeddings of Aoi Ward, Shizuoka City. The color of bars denotes the type of grid indicators and the length of the bar indicates the importance.
  • Figure 5: The performance of different combination for multi-task weights.The task weights of $[0.1,0.1,0.8]$ for city-grid, grid-city and grid-grid commuting flows achieve the best performance.
  • ...and 1 more figures