Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

XTraffic: A Dataset Where Traffic Meets Incidents with Explainability and More

Xiaochuan Gou, Ziyue Li, Tian Lan, Junpeng Lin, Zhishuai Li, Bingyu Zhao, Chen Zhang, Di Wang, Xiangliang Zhang

TL;DR

XTraffic addresses the gap where traffic dynamics and incidents are treated separately by introducing a large-scale, spatiotemporally aligned datacube that links traffic time-series with incident records and road meta-features. The dataset enables four tasks—post-incident traffic forecasting, incident classification from traffic indices, global causal analysis with MM-DAG, and local causal analysis with PCMCI+—and is demonstrated through descriptive statistics, incident-aware forecasting, classification, and both global and local causal inferences. Key findings show that static road attributes meaningfully influence incidents and traffic, while dynamic factors vary by district, and that post-incident data introduce notable forecasting challenges. The resource, released at http://xaitraffic.github.io, promises practical impact for interpretable traffic modeling and policy planning by providing rich, multimodal context and causal insights.

Abstract

Long-separated research has been conducted on two highly correlated tracks: traffic and incidents. Traffic track witnesses complicating deep learning models, e.g., to push the prediction a few percent more accurate, and the incident track only studies the incidents alone, e.g., to infer the incident risk. We, for the first time, spatiotemporally aligned the two tracks in a large-scale region (16,972 traffic nodes) over the whole year of 2023: our XTraffic dataset includes traffic, i.e., time-series indexes on traffic flow, lane occupancy, and average vehicle speed, and incidents, whose records are spatiotemporally-aligned with traffic data, with seven different incident classes. Additionally, each node includes detailed physical and policy-level meta-attributes of lanes. Our data can revolutionalize traditional traffic-related tasks towards higher interpretability and practice: instead of traditional prediction or classification tasks, we conduct: (1) post-incident traffic forecasting to quantify the impact of different incidents on traffic indexes; (2) incident classification using traffic indexes to determine the incidents types for precautions measures; (3) global causal analysis among the traffic indexes, meta-attributes, and incidents to give high-level guidance of the interrelations of various factors; (4) local causal analysis within road nodes to examine how different incidents affect the road segments' relations. The dataset is available at http://xaitraffic.github.io.

XTraffic: A Dataset Where Traffic Meets Incidents with Explainability and More

TL;DR

XTraffic addresses the gap where traffic dynamics and incidents are treated separately by introducing a large-scale, spatiotemporally aligned datacube that links traffic time-series with incident records and road meta-features. The dataset enables four tasks—post-incident traffic forecasting, incident classification from traffic indices, global causal analysis with MM-DAG, and local causal analysis with PCMCI+—and is demonstrated through descriptive statistics, incident-aware forecasting, classification, and both global and local causal inferences. Key findings show that static road attributes meaningfully influence incidents and traffic, while dynamic factors vary by district, and that post-incident data introduce notable forecasting challenges. The resource, released at http://xaitraffic.github.io, promises practical impact for interpretable traffic modeling and policy planning by providing rich, multimodal context and causal insights.

Abstract

Long-separated research has been conducted on two highly correlated tracks: traffic and incidents. Traffic track witnesses complicating deep learning models, e.g., to push the prediction a few percent more accurate, and the incident track only studies the incidents alone, e.g., to infer the incident risk. We, for the first time, spatiotemporally aligned the two tracks in a large-scale region (16,972 traffic nodes) over the whole year of 2023: our XTraffic dataset includes traffic, i.e., time-series indexes on traffic flow, lane occupancy, and average vehicle speed, and incidents, whose records are spatiotemporally-aligned with traffic data, with seven different incident classes. Additionally, each node includes detailed physical and policy-level meta-attributes of lanes. Our data can revolutionalize traditional traffic-related tasks towards higher interpretability and practice: instead of traditional prediction or classification tasks, we conduct: (1) post-incident traffic forecasting to quantify the impact of different incidents on traffic indexes; (2) incident classification using traffic indexes to determine the incidents types for precautions measures; (3) global causal analysis among the traffic indexes, meta-attributes, and incidents to give high-level guidance of the interrelations of various factors; (4) local causal analysis within road nodes to examine how different incidents affect the road segments' relations. The dataset is available at http://xaitraffic.github.io.
Paper Structure (20 sections, 1 equation, 8 figures, 7 tables)

This paper contains 20 sections, 1 equation, 8 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminary of Four Intended Tasks
  4. Related Work of Traffic and Incident Analysis
  5. XTraffic Datacube
  6. Comparison with Existing Datasets
  7. Collection and Construction
  8. Experiments
  9. Descriptive Analysis of XTraffic
  10. Traffic Forecasting after Incidents
  11. Incident Classification
  12. Global Traffic Causal Analysis
  13. Local Causal Analysis for Road Relations
  14. Conclusion and Limitations
  15. Appendix
  16. ...and 5 more sections

Figures (8)

  • Figure 1: Our XTraffic contains (a) traffic data with lane-level meta features, (b) incident data, and (c) their essential yet neglected relations
  • Figure 2: Descriptive Analysis of our XTraffic
  • Figure 3: Visualization of the representation of the time series on the dataset, extracted from the last hidden layer of OmniScaleCNN.
  • Figure 4: Global causal: (a) The learned DAG among meta-features, incidents, and traffic indexes. (b) The factual explanation for selected edges.
  • Figure 5: Case of Local Causal Analysis
  • ...and 3 more figures