Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning to Rank Critical Road Segments via Heterogeneous Graphs with OD Flow Integration

Ming Xu, Jinrong Xiang, Zilong Xie, Xiangfu Meng

TL;DR

<3-5 sentence high-level summary>

Abstract

Existing learning-to-rank methods for road networks often fail to incorporate origin-destination (OD) flows and route information, limiting their ability to model long-range spatial dependencies. To address this gap, we propose HetGL2R, a heterogeneous graph learning framework for ranking road-segment importance. HetGL2R builds a tripartite graph that unifies OD flows, routes, and network topology, and further introduces attribute-guided graphs that elevate node attributes into explicit nodes to model functional similarity. A heterogeneous joint random walk algorithm (HetGWalk) samples both graph types to generate context-rich node sequences. These sequences are encoded with a Transformer to learn embeddings that capture long-range structural dependencies driven by OD demand and route configuration, as well as functional associations derived from attribute similarity. Finally, a listwise ranking strategy with a KL-divergence loss evaluates and ranks segment importance. Experiments on three SUMO-generated simulated networks of different scales show that, against state-of-the-art methods, HetGL2R achieves average improvements of approximately 7.52%, 4.40% and 3.57% in ranking performance.

Learning to Rank Critical Road Segments via Heterogeneous Graphs with OD Flow Integration

TL;DR

<3-5 sentence high-level summary>

Abstract

Existing learning-to-rank methods for road networks often fail to incorporate origin-destination (OD) flows and route information, limiting their ability to model long-range spatial dependencies. To address this gap, we propose HetGL2R, a heterogeneous graph learning framework for ranking road-segment importance. HetGL2R builds a tripartite graph that unifies OD flows, routes, and network topology, and further introduces attribute-guided graphs that elevate node attributes into explicit nodes to model functional similarity. A heterogeneous joint random walk algorithm (HetGWalk) samples both graph types to generate context-rich node sequences. These sequences are encoded with a Transformer to learn embeddings that capture long-range structural dependencies driven by OD demand and route configuration, as well as functional associations derived from attribute similarity. Finally, a listwise ranking strategy with a KL-divergence loss evaluates and ranks segment importance. Experiments on three SUMO-generated simulated networks of different scales show that, against state-of-the-art methods, HetGL2R achieves average improvements of approximately 7.52%, 4.40% and 3.57% in ranking performance.
Paper Structure (25 sections, 15 equations, 8 figures, 10 tables, 1 algorithm)

This paper contains 25 sections, 15 equations, 8 figures, 10 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works
  3. Evaluation of node importance in networks
  4. Learning to rank
  5. Heterogeneous graph learning
  6. Preliminary
  7. Definitions and problem statement
  8. Research objectives
  9. Methodology
  10. HetGWalk algorithm
  11. HetGWalk Overview
  12. Theoretical Properties of HetGWalk
  13. Road segment embedding
  14. Listwise learning to rank module
  15. Computational Complexity Analysis
  16. ...and 10 more sections

Figures (8)

  • Figure 1: A small simulated road network. The network contains 11 nodes and two OD pairs: $<v_1$, $v_4>$ with 300 vehicles on one path, and $<v_8$, $v_7>$ with 500 vehicles on two paths.
  • Figure 2: Framework of the proposed method.
  • Figure 3: Joint random walk sampling procedure.
  • Figure 4: The Performance of Each Method on SY-Net110.
  • Figure 5: The Performance of Each Method on SY-Net514.
  • ...and 3 more figures