Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Reliable and Interpretable Traffic Crash Pattern Prediction and Safety Interventions Using Customized Large Language Models

Yang Zhao, Pu Wang, Yibo Zhao, Hongru Du, Hao Frank Yang

TL;DR

This work introduces TrafficSafe, a framework that converts heterogeneous traffic crash data into textual prompts and leverages a fine-tuned large language model with LoRA for event-level crash outcome prediction. It pairs this predictive capability with TrafficSafe Attribution, a sentence-level Shapley-based framework that enables conditional risk analysis and guides data-collection priorities. The TrafficSafe Event dataset contains 58,903 prompts (~12.74 million words) derived from multi-modal sources, and TrafficSafe LLM delivers a $41.7\%$ average improvement in F1-score over baselines, with high trustworthiness evidenced by confidence-linked accuracy and precision for fatal crashes. This approach offers actionable, interpretable insights for targeted safety interventions under Vision Zero objectives and demonstrates how AI-driven transcription of multi-modal data can support real-world traffic safety policy.

Abstract

Predicting crash events is crucial for understanding crash distributions and their contributing factors, thereby enabling the design of proactive traffic safety policy interventions. However, existing methods struggle to interpret the complex interplay among various sources of traffic crash data, including numeric characteristics, textual reports, crash imagery, environmental conditions, and driver behavior records. As a result, they often fail to capture the rich semantic information and intricate interrelationships embedded in these diverse data sources, limiting their ability to identify critical crash risk factors. In this research, we propose TrafficSafe, a framework that adapts LLMs to reframe crash prediction and feature attribution as text-based reasoning. A multi-modal crash dataset including 58,903 real-world reports together with belonged infrastructure, environmental, driver, and vehicle information is collected and textualized into TrafficSafe Event Dataset. By customizing and fine-tuning LLMs on this dataset, the TrafficSafe LLM achieves a 42% average improvement in F1-score over baselines. To interpret these predictions and uncover contributing factors, we introduce TrafficSafe Attribution, a sentence-level feature attribution framework enabling conditional risk analysis. Findings show that alcohol-impaired driving is the leading factor in severe crashes, with aggressive and impairment-related behaviors having nearly twice the contribution for severe crashes compared to other driver behaviors. Furthermore, TrafficSafe Attribution highlights pivotal features during model training, guiding strategic crash data collection for iterative performance improvements. The proposed TrafficSafe offers a transformative leap in traffic safety research, providing a blueprint for translating advanced AI technologies into responsible, actionable, and life-saving outcomes.

Towards Reliable and Interpretable Traffic Crash Pattern Prediction and Safety Interventions Using Customized Large Language Models

TL;DR

This work introduces TrafficSafe, a framework that converts heterogeneous traffic crash data into textual prompts and leverages a fine-tuned large language model with LoRA for event-level crash outcome prediction. It pairs this predictive capability with TrafficSafe Attribution, a sentence-level Shapley-based framework that enables conditional risk analysis and guides data-collection priorities. The TrafficSafe Event dataset contains 58,903 prompts (~12.74 million words) derived from multi-modal sources, and TrafficSafe LLM delivers a average improvement in F1-score over baselines, with high trustworthiness evidenced by confidence-linked accuracy and precision for fatal crashes. This approach offers actionable, interpretable insights for targeted safety interventions under Vision Zero objectives and demonstrates how AI-driven transcription of multi-modal data can support real-world traffic safety policy.

Abstract

Predicting crash events is crucial for understanding crash distributions and their contributing factors, thereby enabling the design of proactive traffic safety policy interventions. However, existing methods struggle to interpret the complex interplay among various sources of traffic crash data, including numeric characteristics, textual reports, crash imagery, environmental conditions, and driver behavior records. As a result, they often fail to capture the rich semantic information and intricate interrelationships embedded in these diverse data sources, limiting their ability to identify critical crash risk factors. In this research, we propose TrafficSafe, a framework that adapts LLMs to reframe crash prediction and feature attribution as text-based reasoning. A multi-modal crash dataset including 58,903 real-world reports together with belonged infrastructure, environmental, driver, and vehicle information is collected and textualized into TrafficSafe Event Dataset. By customizing and fine-tuning LLMs on this dataset, the TrafficSafe LLM achieves a 42% average improvement in F1-score over baselines. To interpret these predictions and uncover contributing factors, we introduce TrafficSafe Attribution, a sentence-level feature attribution framework enabling conditional risk analysis. Findings show that alcohol-impaired driving is the leading factor in severe crashes, with aggressive and impairment-related behaviors having nearly twice the contribution for severe crashes compared to other driver behaviors. Furthermore, TrafficSafe Attribution highlights pivotal features during model training, guiding strategic crash data collection for iterative performance improvements. The proposed TrafficSafe offers a transformative leap in traffic safety research, providing a blueprint for translating advanced AI technologies into responsible, actionable, and life-saving outcomes.
Paper Structure (33 sections, 4 equations, 16 figures, 1 table)

This paper contains 33 sections, 4 equations, 16 figures, 1 table.

Table of Contents

  1. Introduction
  2. Novelties and Contributions
  3. Results
  4. Multi-modal Crash Data
  5. TrafficSafe Crash Outcomes Prediction Pipeline
  6. Constructing Prompts and Prediction Targets
  7. Adapting LLM for Crash Prediction
  8. Performance Evaluation of TrafficSafe LLM
  9. TrafficSafe Attribution and Result Interpretation
  10. Factor Attribution at Inference Stage for Conditional Risk Analysis
  11. Factor Attribution at Training Stage for Effective Data Collection and Model Development
  12. Discussion
  13. Methods
  14. TrafficSafe Event Dataset Construction
  15. Raw Data
  16. ...and 18 more sections

Figures (16)

  • Figure 1: Overview of the Proposed TrafficSafe Framework. (a) The U.S. faces one of the highest crash risks among developed countries, with a rising trend. However, analyzing and addressing this issue is challenging due to the heterogeneous factors involved in crash events, including traffic conditions, human behavior, environmental impacts, and driver characteristics. To tackle this, we propose TrafficSafe, a framework designed for two key tasks: 1) Predicting crash outcomes and 2) Attributing crash factors with conditional risk analysis. By addressing questions such as why crashes occur and how to mitigate crash risks, TrafficSafe seeks to deliver optimal policy for safety improvement, aligning with the Vision Zero goal fhwa_vision_zero. (b) The TrafficSafe workflow incorporates multi-modal data, including driver behavior, vehicle details, infrastructure, and environmental conditions, represented through textual reports, satellite imagery, and other formats. Leveraging an AI-expert cooperative method, the crash data is transformed into textual prompts, resulting in the TrafficSafe Event dataset comprising 58,903 prompts. TrafficSafe LLM is created with accurate and trustworthy forecasting abilities for further analysis. Building on this pipeline, TrafficSafe Attribution operates across three dimensions: 1) Event-level risk analysis to identify feature contributions, 2) Conditional risk analysis to assess state-level risks under varying conditions, and 3) Data collection guidance to optimize the data acquisition process. The results of TrafficSafe Attribution provide actionable insights to enhance data analysis and collection, fostering a more comprehensive understanding of crash data and events.
  • Figure 1: A Crash Event Prompt Example from Washington Dataset.
  • Figure 2: TrafficSafe Crash Outcomes Prediction Pipeline. Multi-modal crash data is collected and organized into textual prompts through an AI-expert cooperative process. The HSIS crash data, satellite images, and infrastructure data are used to extract general and infrastructure information, including the crash time, location, the road level, and so on. The vehicle data, and person data are converted into the event information and the unit information, including vehicle movements, driver characteristics (e.g., age, gender, alcohol use), vehicle attributes (e.g., manufacture year), and so on. TrafficSafe Event dataset is created with three prediction targets: Injury, Severity, and Type. The Injury task predicts the number of people injured in the crash event, the Severity task estimates the severity level of the crash, such as no apparent injury or fatal, and the Type task classifies type of crash, such as single vehicle with object or angle impacts right (The crash event consequences classification are provided in Supplementary Table 2 and Supplementary Table 3. The TrafficSafe LLM is fine-tuned using the TrafficSafe Event dataset. To reframe the crash outcomes prediction from a classification task to a language inference task, TrafficSafe LLM is fine-tuned by adding prediction targets as special tokens in its vocabulary and adjusting parameters using Low-Rank Adaptations (LoRA) hu2021lora.
  • Figure 2: A Crash Event Prompt Prompt Example from Illinois Dataset.
  • Figure 3:
  • ...and 11 more figures