DrivingResearch.

A conflict-based approach for real-time road safety analysis: Comparative evaluation with crash-based models

Orsini, Federico; Gecchele, Gregorio; Rossi, Riccardo; Gastaldi, Massimiliano · 2021 · Accident Analysis & Prevention

DOI: 10.1016/j.aap.2021.106382

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the limitations of traditional real-time crash prediction models (RTCPMs), which rely on crash data that is often scarce, unreliable, or lacks precise spatial and temporal accuracy. To overcome these issues, the authors propose a real-time conflict prediction model (RTConfPM) that uses surrogate measures of safety—specifically traffic conflicts—as precursors to crashes. The study aims to determine whether predicting unsafe situations based on conflicts offers superior performance compared to traditional crash-based models, particularly in scenarios where high-quality crash data is unavailable. The methodology involves developing and comparing an RTConfPM and an RTCPM using data collected from a 150-km Italian motorway over one year. Microwave radar sensors at 40 cross-sections recorded vehicle-by-vehicle data, which was aggregated into 5-minute intervals. Input variables included traffic volume, heavy-duty vehicle percentage, harmonic mean speed, and speed variance for each of the three lanes. For the RTConfPM, unsafe situations were defined using time-to-collision (TTC) values; a statistical analysis based on extreme value theory established a TTC threshold of 0.78 seconds, with an interval classified as unsafe if at least three such conflicts occurred. Both models utilized Random Forest for variable selection, the Synthetic Minority Oversampling Technique (SMOTE) to balance the highly skewed datasets, and Support Vector Machines (SVM) as the classifier. Model robustness was assessed using Monte Carlo cross-validation. The results demonstrate that the conflict-based approach significantly outperforms the traditional crash-based model. The RTConfPM achieved more than 93% accuracy, recall, and specificity in predicting unsafe situations within a 5-minute window. In contrast, the RTCPM performed poorly due to the rarity and inconsistency of crash data relative to traffic conditions. The study highlights that traffic conditions leading to conflicts are consistent precursors to risk, whereas crashes are stochastic outcomes influenced by driver interventions, making conflict data a more reliable training signal for real-time prediction. The significance of this work lies in providing a viable alternative for real-time road safety analysis that does not depend on the availability of precise crash records. By leveraging frequent conflict data derived from standard radar infrastructure, the RTConfPM offers a robust tool for identifying risky traffic conditions. This approach facilitates timely interventions, such as variable message signs or ramp metering, to prevent crashes before they occur. The findings suggest that conflict-based models are more reliable for real-time applications and have strong potential for large-scale implementation in intelligent transportation systems.

Key finding

The real-time conflict prediction model significantly outperformed the traditional crash-based model, achieving over 93% accuracy, recall, and specificity in predicting unsafe traffic situations.

Methodology

naturalistic

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via author_sweep_intake on 2026-05-28.

StageOutcomeToolModelPromptAttemptsCompleted
discover success author_sweep — — 2 2026-05-28
archive success canonical_url — — 1 2026-06-06
extract success cached — — 3 2026-06-10
clean success clean — — 1 2026-06-07
chunk success chunk — — 1 2026-06-07
embed success embed Qwen/Qwen3-Embedding-8B — 1 2026-06-07
enrich failed — — — 5 2026-07-02
promote success — — — 1 2026-06-04
summarize success llm qwen3.6-27b-prismaquant summ-v5 2 2026-06-10
tag success vector_similarity — — 15 2026-06-11
verify success — — — 2 2026-06-10

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).