Decision-adjusted driver risk predictive models using kinematics information
DOI: 10.1016/j.aap.2021.106088
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 challenge of accurately predicting individual driver crash risk, a task complicated by the rarity of crash events and significant heterogeneity among drivers. While telematics data from connected vehicle technology offers high-resolution kinematic information (e.g., acceleration, deceleration, lateral movement) that could improve risk prediction, existing methods often rely on subjective threshold selections for defining risky kinematic events or generic model selection criteria like the Area Under the Curve (AUC). These generic approaches may not optimize performance for specific decision goals, such as identifying the top percentage of highest-risk drivers for targeted safety interventions. The authors propose a "decision-adjusted" modeling framework to systematically determine optimal kinematic thresholds and model parameters tailored to specific decision rules. The researchers applied this framework to data from the Second Strategic Highway Research Program (SHRP 2) Naturalistic Driving Study, which included 3,440 drivers and 1,161,112 driving hours. The dataset provided detailed kinematic measurements (longitudinal and lateral acceleration) alongside traditional predictors such as demographics, personality factors, and self-reported driving history. The study defined high G-force events based on acceleration (ACC), deceleration (DEC), and lateral acceleration (LAT) exceeding specific thresholds. The decision-adjusted model was optimized to maximize prediction precision—defined as the percentage of correctly identified high-risk drivers—across various decision rules targeting the top 1% to 20% of riskiest drivers. This approach was compared against a baseline model using only non-telematics predictors and a model optimized using the standard AUC criterion. The results demonstrated that the decision-adjusted model significantly outperformed the baseline model, improving prediction precision by 6.3% to 26.1%. Furthermore, it surpassed the AUC-optimized model with precision improvements ranging from 5.3% to 31.8%. The study found that optimal thresholds for ACC, DEC, and LAT were highly sensitive to the specific decision rule applied, particularly when targeting a small percentage of high-risk drivers. This sensitivity confirms that a one-size-fits-all threshold for kinematic events is suboptimal and that model parameters must be adjusted to align with the specific operational goal of the risk prediction system. The significance of this work lies in its demonstration of the value of kinematic driving behavior in crash risk prediction and the necessity of a systematic, objective-driven approach to feature engineering. By linking model optimization directly to specific decision goals, such as fleet safety management or use-based insurance, the proposed framework allows for more effective identification of high-risk drivers. This enables better allocation of limited resources for safety countermeasures and driver behavior interventions. The study concludes that leveraging telematics data through decision-adjusted modeling provides a superior alternative to traditional methods, offering a robust tool for developing connected-vehicle safety technologies and targeted risk mitigation strategies.
Key finding
The decision-adjusted modeling framework improved driver risk prediction precision by 6.3% to 26.1% compared to baseline non-telematics models and outperformed models optimized using receiver operating characteristic curve criteria.
Methodology
naturalistic
Sample size: 3440
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 | — | — | — | 1 | 2026-05-28 |
| archive | success | canonical_url | — | — | 4 | 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 | success | openalex | — | — | 4 | 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 | partial | — | — | — | 2 | 2026-06-10 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified_with_issues.
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
- Theoretical Contribution: computational model