Analyzing Pile-Up Crash Severity: Insights from Real-Time Traffic and Environmental Factors Using Ensemble Machine Learning and Shapley Additive Explanations Method
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This study addresses the under-researched problem of pile-up (PU) crash severity, defined as collisions involving three or more vehicles within a brief timeframe. While PU crashes account for a small fraction of total incidents, they result in disproportionate fatalities, injuries, and economic losses. The research aims to identify the specific real-time traffic and environmental factors, as well as their interactions, that contribute to severe PU crashes (resulting in fatality or injury) versus property damage-only crashes. The study is motivated by the limitations of previous research, which often relied on static traffic measures like Annual Average Daily Traffic (AADT) and lacked detailed analysis of heavy vehicle proportions and complex factor interactions. To achieve this, the authors analyzed 2,165 PU crashes recorded on eleven major suburban freeways in Iran between March 2014 and March 2019. The dataset combined police-reported crash characteristics with real-time traffic data from loop detectors, specifically focusing on the one-hour period preceding each crash. Key variables included the Total Volume/Capacity (TV/C) ratio, the ratio of Heavy Vehicles Volume to Total Volume (HVV/TV), and average speed. The researchers employed six machine learning models to predict crash severity, selecting the CatBoost model for its superior performance. To interpret the complex "black-box" model outputs and identify critical thresholds and interactions, the study utilized the Shapley Additive Explanations (SHAP) method, which allows for the estimation of individual variable effects and their combined influences. The results indicate that several specific conditions significantly increase the risk of severe PU crashes. High average speeds (>90 km/h), low congestion levels (TV/C < 0.6), and a higher proportion of heavy vehicles (HVV/TV ≥ 0.1) were identified as primary risk factors. Environmental and geometric factors, including horizontal curves, longitudinal grades, nighttime conditions, and the direct involvement of heavy vehicles, also correlated with increased severity. Crucially, the SHAP analysis revealed significant interaction effects. Severe crashes were particularly associated with the co-occurrence of very low congestion (TV/C ≈ 0.1), high heavy vehicle presence (HVV/TV ≥ 0.25), and nighttime driving. Additionally, interactions between low-to-moderate congestion (TV/C ≈ 0.1 or 0.45), high heavy vehicle ratios, and high speeds (>90 km/h) were found to exacerbate crash severity. The significance of this research lies in its provision of precise, data-driven insights into the dynamics of severe pile-up crashes, moving beyond aggregated traffic metrics to real-time conditions. By identifying critical thresholds for traffic saturation and heavy vehicle presence, the findings support more informed decision-making for transportation policymakers and safety planners. The study highlights that less congested conditions, often assumed to be safer, can pose higher risks for severe PU crashes when combined with high speeds and heavy vehicle involvement. These insights can guide the development of targeted countermeasures, such as dynamic speed management and improved monitoring of heavy vehicle traffic during nighttime hours, to mitigate the severe consequences of pile-up incidents.
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-06-18 |
| archive | success | core_acuk | — | — | 3 | 2026-06-25 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-19 |
| chunk | success | chunk | — | — | 1 | 2026-06-19 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-19 |
| promote | success | — | — | — | 1 | 2026-06-18 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-19 |
| verify | success | — | — | — | 1 | 2026-06-26 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; 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).
- Empirical Findings: crash risk outcomes
- Methodological Resource: dataset resource