pith. machine review for the scientific record. sign in

arxiv: 2605.06684 · v1 · submitted 2026-04-25 · 💻 cs.LG

Recognition: no theorem link

From Canopy to Collision: A Hybrid Predictive Framework for Identifying Risk Factors in Tree-Involved Traffic Crashes

Abdul Azim , Ahmed Hossain , Soumyadip Maitra , Panick Kalambay
Authors on Pith no claims yet

Pith reviewed 2026-05-11 00:54 UTC · model grok-4.3

classification 💻 cs.LG
keywords tree-involved crashescrash severityrestraint non-useCatBoostSHAP explanationsrisk factorsrun-off-road collisionslogistic regression
0
0 comments X

The pith

Restraint non-use emerges as the dominant risk factor for severe injury in tree-involved crashes, with unrestrained occupants nearly three times more likely to suffer fatal or incapacitating outcomes.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper builds a multi-step modeling pipeline on four years of U.S. crash records to isolate the factors that turn a run-off-road collision with a tree into a severe injury event. It shows that failure to use restraints stands out above all other variables, largely because it allows ejection during the high-energy impact. Vehicle age, speeding, and driver impairment also raise severity, and certain combinations of conditions such as darkness with older vehicles produce added risk. The work validates its machine-learning rankings with traditional regression and interaction plots to strengthen the evidence for targeted safety actions.

Core claim

A hybrid framework first trains a CatBoost classifier on binary injury severity (fatal or incapacitating versus lesser injury) using Crash Report Sampling System data from 2020-2023, then applies SHAP values to rank feature influence, fits a logistic regression model to confirm effect sizes, and generates interaction plots. This process identifies restraint non-use as the strongest predictor, with unrestrained occupants facing approximately three times higher odds of severe injury due to ejection risk; vehicle age, speeding violations, and driver impairment each exert substantial additional effects; and specific pairwise interactions, including lighting with vehicle age and speeding with low

What carries the argument

Hybrid predictive framework that combines CatBoost classification, SHAP value ranking, logistic regression validation, and interaction analysis applied to national crash sampling data to quantify and explain severity risk factors.

If this is right

  • Unrestrained occupants face nearly three times the risk of severe injury in tree crashes, driven primarily by ejection.
  • Older vehicles, speeding violations, and driver impairment each produce large increases in the probability of severe outcomes.
  • Interactions between lighting conditions and vehicle age, speeding and lighting, restraint use and vehicle age, and road surface and speeding create additive risk elevations.
  • The results support safety interventions focused on seatbelt enforcement, speed management in reduced visibility, and replacement of older vehicles in the fleet.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar hybrid modeling could be applied to other fixed-object crash types to test whether restraint non-use remains the leading factor across different impact energies.
  • Linking these findings to regional enforcement data could quantify how much severe injury reduction would follow from higher restraint compliance rates.
  • Incorporating vehicle telemetry on actual restraint use and speed at impact would allow direct testing of the ejection mechanism in future datasets.

Load-bearing premise

The models assume that the statistical associations they detect, such as between restraint non-use and injury severity, represent causal risk relationships rather than correlations, and that the sampled crash reports accurately represent the full population of tree-involved incidents without major bias or underreporting.

What would settle it

A before-and-after comparison of severe-injury rates in tree-involved crashes in a jurisdiction that implements a sustained seatbelt enforcement program, or a matched simulation study that enforces restraint use while holding other factors constant and measures the resulting change in injury outcomes.

read the original abstract

Tree-involved crashes represent a critical subset of run-off-road (ROR) collisions, often resulting in fatal or severe injuries due to high-energy impacts. This study develops a comprehensive analytical framework to identify and quantify risk factors contributing to crash severity in tree-involved collisions using the Crash Report Sampling System (CRSS) database spanning 2020-2023. The modeling framework follows a multi-step process. First, a machine learning based classification model (CatBoost) identifies key factors associated with binary crash injury severity (KA: fatal or incapacitating injury versus BC: non-incapacitating or possible injury). Second, SHapley Additive exPlanations (SHAP) tool is used to quantify and visualize the marginal effects of top influential factors on crash severity. Third, a binary logistic regression model estimates factor effects and validates SHAP-derived importance measures. Finally, SHAP interaction plots examine the combined effects of key contributing factors. Results reveal restraint non-use as the most influential predictor, with unrestrained occupants nearly three times more likely to experience severe outcomes due to ejection risk. Vehicle age, speeding violations, and driver impairment demonstrate substantial effects, reflecting reduced crashworthiness, increased impact forces, and reduced control capabilities. Critical interactions emerge between lighting conditions and vehicle age, speeding and lighting conditions, restraint use and vehicle age, and road surface and speeding, demonstrating additive risk effects with specific interactions. These findings provide critical insights for targeted safe system-based interventions, including enhanced seat belt enforcement, speed management in reduced visibility conditions, and vehicle fleet modernization.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper develops a hybrid analytical framework using CatBoost classification, SHAP explanations, binary logistic regression, and interaction plots to identify risk factors for binary injury severity (KA vs. BC) in tree-involved crashes from the CRSS 2020-2023 database. It reports restraint non-use as the dominant predictor (unrestrained occupants nearly three times more likely to experience severe outcomes), with substantial effects from vehicle age, speeding violations, and driver impairment, plus key interactions (e.g., lighting and vehicle age).

Significance. If properly validated, the work could add to traffic safety literature by applying interpretable ML to a focused crash subset and highlighting modifiable factors for interventions such as seat-belt enforcement and speed management in low-visibility conditions. The multi-method design (ML importance plus regression cross-check) is a reasonable approach for observational crash data, though the absence of performance metrics and causal safeguards limits its current contribution.

major comments (3)
  1. [Abstract and Modeling Framework] Abstract and Modeling Framework section: the multi-step process is outlined but no model performance metrics (accuracy, AUC, F1, confusion matrix), no sample size for the tree-involved subset, and no validation details (cross-validation, train-test split, or hyperparameter tuning procedure) are reported. This is load-bearing for the central claim because SHAP importance rankings and the subsequent logistic regression coefficients cannot be interpreted without evidence that the CatBoost model has adequate predictive power on the data.
  2. [Abstract and Results] Abstract and Results section: the statement that unrestrained occupants are 'nearly three times more likely to experience severe outcomes due to ejection risk' and parallel mechanistic attributions ('reduced crashworthiness' for vehicle age, 'increased impact forces' for speeding) treat observational associations as causal. The logistic regression and SHAP values are derived from police-reported CRSS data without DAGs, propensity methods, instrumental variables, or sensitivity analyses for unmeasured confounding (driver behavior, crash circumstances). This directly undermines the risk-factor interpretations.
  3. [Modeling Framework and Results] Modeling Framework and Results: logistic regression is positioned as validation for SHAP-derived importance on the identical fitted dataset, introducing circularity. No independent hold-out set, temporal validation, or external benchmark dataset is described, weakening the cross-check claim.
minor comments (2)
  1. Include a table of descriptive statistics for the analytic sample (number of tree-involved crashes, severity distribution, missing-data handling) to allow readers to assess generalizability.
  2. [Methods] Clarify the exact CRSS variable definitions and coding for 'KA' vs. 'BC' severity and for restraint use in the methods section.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which help strengthen the manuscript. We address each major comment point by point below, agreeing with the need for greater transparency and precision where warranted, and outlining specific revisions.

read point-by-point responses

  1. Referee: [Abstract and Modeling Framework] Abstract and Modeling Framework section: the multi-step process is outlined but no model performance metrics (accuracy, AUC, F1, confusion matrix), no sample size for the tree-involved subset, and no validation details (cross-validation, train-test split, or hyperparameter tuning procedure) are reported. This is load-bearing for the central claim because SHAP importance rankings and the subsequent logistic regression coefficients cannot be interpreted without evidence that the CatBoost model has adequate predictive power on the data.

    Authors: We agree that these elements are essential for evaluating model reliability and supporting the SHAP and logistic regression interpretations. The manuscript as submitted omitted them. In revision we will report the exact sample size of the tree-involved CRSS subset, include standard performance metrics (AUC, accuracy, F1-score, precision, recall, and confusion matrix) for the CatBoost model, and describe the validation protocol including any train-test split, cross-validation folds, and hyperparameter tuning procedure. These additions will directly address the concern about predictive power. revision: yes

  2. Referee: [Abstract and Results] Abstract and Results section: the statement that unrestrained occupants are 'nearly three times more likely to experience severe outcomes due to ejection risk' and parallel mechanistic attributions ('reduced crashworthiness' for vehicle age, 'increased impact forces' for speeding) treat observational associations as causal. The logistic regression and SHAP values are derived from police-reported CRSS data without DAGs, propensity methods, instrumental variables, or sensitivity analyses for unmeasured confounding (driver behavior, crash circumstances). This directly undermines the risk-factor interpretations.

    Authors: The referee is correct that certain phrasing in the abstract and results implies causal mechanisms. Because the analysis is observational, we will revise all such language to emphasize associations (e.g., “associated with nearly three times higher odds of severe injury, consistent with ejection risk”) and will qualify mechanistic statements as hypothesized explanations drawn from prior literature rather than direct inferences from the data. We will also add an explicit limitations paragraph discussing the absence of causal identification strategies and the possibility of unmeasured confounding. revision: yes

  3. Referee: [Modeling Framework and Results] Modeling Framework and Results: logistic regression is positioned as validation for SHAP-derived importance on the identical fitted dataset, introducing circularity. No independent hold-out set, temporal validation, or external benchmark dataset is described, weakening the cross-check claim.

    Authors: We acknowledge that applying logistic regression to the same observations used for CatBoost introduces dependence and that the term “validation” was imprecise. The logistic model was intended as a complementary, easily interpretable check on the direction and ranking of effects identified by SHAP. In revision we will reframe this section to describe the logistic regression as a complementary interpretability tool rather than independent validation, explicitly note the shared data as a limitation, and explore whether a temporal split (e.g., 2020–2022 training, 2023 testing) is feasible with the CRSS years available. We will also discuss the scarcity of external benchmark datasets for this narrow crash type. revision: partial

Circularity Check

0 steps flagged

No circularity: standard empirical ML + stats pipeline on observational data

full rationale

The paper applies CatBoost for classification, SHAP for feature importance, and logistic regression for coefficient estimation and cross-method validation, all fitted to the same CRSS 2020-2023 sample of tree-involved crashes. No derivation chain is claimed from first principles; results are presented as associations derived from the fitted models. Logistic regression is used to estimate effects and compare importance rankings to SHAP, but this is cross-validation of two models on shared data rather than a prediction that reduces to its inputs by construction. No self-definitional steps, self-citation load-bearing premises, or renamed known results are present. The framework is self-contained against external benchmarks only in the sense that it reports data-driven patterns without claiming causal identification or out-of-sample prediction beyond the sample.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The framework relies on standard assumptions in observational crash data analysis and machine learning without introducing new physical entities; free parameters are implicit in model fitting.

free parameters (2)
  • CatBoost hyperparameters
    Tuned for binary classification on crash data but specific values and tuning process not detailed in abstract.
  • Logistic regression coefficients
    Estimated from data to quantify marginal effects and validate SHAP rankings.
axioms (2)
  • domain assumption CRSS database provides a representative sample of tree-involved crashes without major selection bias
    All modeling and conclusions rest on this 2020-2023 dataset as the sole empirical source.
  • domain assumption Binary severity classification (KA vs BC) accurately reflects true injury outcomes
    Serves as the target variable for both CatBoost and logistic models.

pith-pipeline@v0.9.0 · 5596 in / 1588 out tokens · 74176 ms · 2026-05-11T00:54:34.745790+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

82 extracted references · 70 canonical work pages

  1. [1]

    Department of Transportation, 2021

    Federal Highway Administration (FHWA), Tree Crashes, U.S. Department of Transportation, 2021. https://highways.dot.gov/sites/fhwa.dot.gov/files/FHWA-SA-21-022_Tree_Crashes.pdf

    2021

  2. [2]

    Bucsuházy, R

    K. Bucsuházy, R. Zůvala, V. Valentová, J. Ambros, Factors related to severe single-vehicle tree crashes: In-depth crash study, PLOS ONE 17 (2022) e0248171. https://doi.org/10.1371/journal.pone.0248171

    work page doi:10.1371/journal.pone.0248171 2022

  3. [3]

    Pintar, N

    F.A. Pintar, N. Yoganandan, D.J. Maiman, Injury mechanisms and severity in narrow offset frontal impacts, Ann. Adv. Automot. Med. Assoc. Adv. Automot. Med. Annu. Sci. Conf. 52 (2008) 185– 189

    2008

  4. [4]

    https://austroads.gov.au/publications/road-design/agrd06

    Austroad, Guide to Road Design Part 6: Roadside Design, Safety and Barriers, Austroads, 2024. https://austroads.gov.au/publications/road-design/agrd06

    2024

  5. [5]

    Pintar, D.J

    F.A. Pintar, D.J. Maiman, N. Yoganandan, Injury Patterns in Side Pole Crashes, Annu. Proc. Assoc. Adv. Automot. Med. 51 (2007) 419–433. https://pmc.ncbi.nlm.nih.gov/articles/PMC3217499/

    2007

  6. [6]

    Turner, E.R

    D.S. Turner, E.R. Mansfield, Urban Trees and Roadside Safety, J. Transp. Eng. 116 (1990) 90–104. https://doi.org/10.1061/(ASCE)0733-947X(1990)116:1(90)

    work page doi:10.1061/(asce)0733-947x(1990)116:1(90 1990

  7. [7]

    Holdridge, V.N

    J.M. Holdridge, V.N. Shankar, G.F. Ulfarsson, The crash severity impacts of fixed roadside objects, J. Safety Res. 36 (2005) 139–147. https://doi.org/10.1016/j.jsr.2004.12.005

    work page doi:10.1016/j.jsr.2004.12.005 2005

  8. [8]

    Yamamoto, V.N

    T. Yamamoto, V.N. Shankar, Bivariate ordered-response probit model of driver’s and passenger’s injury severities in collisions with fixed objects, Accid. Anal. Prev. 36 (2004) 869–876. https://doi.org/10.1016/j.aap.2003.09.002

    work page doi:10.1016/j.aap.2003.09.002 2004

  9. [9]

    Daniello, H.C

    A. Daniello, H.C. Gabler, Fatality risk in motorcycle collisions with roadside objects in the United States, Accid. Anal. Prev. 43 (2011) 1167–1170. https://doi.org/10.1016/j.aap.2010.12.027

    work page doi:10.1016/j.aap.2010.12.027 2011

  10. [10]

    Bambach, R.H

    M.R. Bambach, R.H. Grzebieta, J. Olivier, A.S. McIntosh, Fatality Risk for Motorcyclists in Fixed Object Collisions, J. Transp. Saf. Secur. 3 (2011) 222–235. https://doi.org/10.1080/19439962.2011.587940

    work page doi:10.1080/19439962.2011.587940 2011

  11. [11]

    Shrestha, L

    R. Shrestha, L. Ventura, N. Venkataraman, V. Shankar, An error components mixed logit with heterogeneity in means and variance for fixed object occupant severity outcomes, Anal. Methods Accid. Res. 42 (2024) 100330. https://doi.org/10.1016/j.amar.2024.100330

    work page doi:10.1016/j.amar.2024.100330 2024

  12. [12]

    Rahman, S

    Md.M. Rahman, S. Hernandez, R.M. Radwan Albatayneh, Assessing the impact of COVID-19 on driver injury severities in fixed-object passenger car crashes: Insights from temporal and partially constrained modeling analysis, Anal. Methods Accid. Res. 47 (2025) 100397. https://doi.org/10.1016/j.amar.2025.100397

    work page doi:10.1016/j.amar.2025.100397 2025

  13. [13]

    Bendigeri, Analysis of factors contributing to roadside tree crashes in South Carolina, Theses (2009)

    V. Bendigeri, Analysis of factors contributing to roadside tree crashes in South Carolina, Theses (2009). https://open.clemson.edu/all_theses/711

    2009

  14. [14]

    S. Das, B. Storey, T.H. Shimu, S. Mitra, M. Theel, B. Maraghehpour, Severity analysis of tree and utility pole crashes: Applying fast and frugal heuristics, IATSS Res. 44 (2020) 85–93. https://doi.org/10.1016/j.iatssr.2019.08.001

    work page doi:10.1016/j.iatssr.2019.08.001 2020

  15. [15]

    K.L. Wolf, N. Bratton, Urban Trees and Traffic Safety: Considering U.S. Roadside Policy and Crash Data, Arboric. Urban For. AUF 32 (2006) 170. https://doi.org/10.48044/jauf.2006.023

    work page doi:10.48044/jauf.2006.023 2006

  16. [16]

    Marshall, Y

    W.E. Marshall, Y. Golombek, N. Coppola, B. Janson, The Unresolved Relationship between Street Trees and Road Safety, Mountain-Plains Consortium, Fargo, ND, 2019. https://trid.trb.org/View/1650959

    work page arXiv 2019

  17. [17]

    Van Treese Ii, A

    J. Van Treese Ii, A. Koeser, G. Fitzpatrick, M. Olexa, E. Allen, Frequency and Severity of Tree and Other Fixed Object Crashes in Florida, 2006—2013, Arboric. Urban For. 45 (2019). https://doi.org/10.48044/jauf.2019.006

    work page doi:10.48044/jauf.2019.006 2006

  18. [18]

    Ray, Impact conditions in side-impact collisions with fixed roadside objects, Accid

    M.H. Ray, Impact conditions in side-impact collisions with fixed roadside objects, Accid. Anal. Prev. 31 (1999) 21–30. https://doi.org/10.1016/S0001-4575(98)00041-4

    work page doi:10.1016/s0001-4575(98)00041-4 1999

  19. [19]

    M.H. Ray, K. Hiranmayee, Evaluating Human Risk in Side Impact Collisions with Roadside Objects, Transp. Res. Rec. 1720 (2000) 67–71. https://doi.org/10.3141/1720-08. 27

    work page doi:10.3141/1720-08 2000

  20. [20]

    Naing, J

    C.L. Naing, J. Hill, R. Thomson, H. Fagerlind, M. Kelkka, C. Klootwijk, G. Dupre, O. Bisson, Single-vehicle collisions in Europe: analysis using real-world and crash-test data, Int. J. Crashworthiness 13 (2008) 219–229. https://doi.org/10.1080/13588260701788583

    work page doi:10.1080/13588260701788583 2008

  21. [21]

    Fitzpatrick, The Effect of Roadside Elements on Driver Behavior and Run-Off-the-Road Crash Severity, Doctoral Dissertation, University of Massachusetts Amherst, 2013

    C.D. Fitzpatrick, The Effect of Roadside Elements on Driver Behavior and Run-Off-the-Road Crash Severity, Doctoral Dissertation, University of Massachusetts Amherst, 2013. https://scholarworks.umass.edu/dissertations/AAI3603093

    2013

  22. [22]

    Van Treese II, A.K

    J.W. Van Treese II, A.K. Koeser, G.E. Fitzpatrick, M.T. Olexa, E.J. Allen, A review of the impact of roadway vegetation on drivers’ health and well-being and the risks associated with single-vehicle crashes, Arboric. J. 39 (2017) 179–193. https://doi.org/10.1080/03071375.2017.1374591

    work page doi:10.1080/03071375.2017.1374591 2017

  23. [23]

    Cheng, R

    G. Cheng, R. Cheng, Y. Pei, L. Xu, W. Qi, Severity assessment of accidents involving roadside trees based on occupant injury analysis, PLOS ONE 15 (2020) e0231030. https://doi.org/10.1371/journal.pone.0231030

    work page doi:10.1371/journal.pone.0231030 2020

  24. [24]

    Malyshkina, F.L

    N.V. Malyshkina, F.L. Mannering, Markov switching multinomial logit model: An application to accident-injury severities, Accid. Anal. Prev. 41 (2009) 829–838. https://doi.org/10.1016/j.aap.2009.04.006

    work page doi:10.1016/j.aap.2009.04.006 2009

  25. [25]

    Kockelman, Y.-J

    K.M. Kockelman, Y.-J. Kweon, Driver injury severity: an application of ordered probit models, Accid. Anal. Prev. 34 (2002) 313–321. https://doi.org/10.1016/S0001-4575(01)00028-8

    work page doi:10.1016/s0001-4575(01)00028-8 2002

  26. [26]

    F. Wei, G. Lovegrove, An empirical tool to evaluate the safety of cyclists: Community based, macro-level collision prediction models using negative binomial regression, Accid. Anal. Prev. 61 (2013) 129–137. https://doi.org/10.1016/j.aap.2012.05.018

    work page doi:10.1016/j.aap.2012.05.018 2013

  27. [27]

    Al-Ghamdi, Using logistic regression to estimate the influence of accident factors on accident severity, Accid

    A.S. Al-Ghamdi, Using logistic regression to estimate the influence of accident factors on accident severity, Accid. Anal. Prev. 34 (2002) 729–741. https://doi.org/10.1016/S0001-4575(01)00073-2

    work page doi:10.1016/s0001-4575(01)00073-2 2002

  28. [28]

    Kononen, C.A.C

    D.W. Kononen, C.A.C. Flannagan, S.C. Wang, Identification and validation of a logistic regression model for predicting serious injuries associated with motor vehicle crashes, Accid. Anal. Prev. 43 (2011) 112–122. https://doi.org/10.1016/j.aap.2010.07.018

    work page doi:10.1016/j.aap.2010.07.018 2011

  29. [29]

    Y. Li, R. Gu, J. Lee, M. Yang, Q. Chen, Y. Zhang, The dynamic tradeoff between safety and efficiency in discretionary lane-changing behavior: A random parameters logit approach with heterogeneity in means and variances, Accid. Anal. Prev. 153 (2021) 106036. https://doi.org/10.1016/j.aap.2021.106036

    work page doi:10.1016/j.aap.2021.106036 2021

  30. [30]

    Zhang, N.N

    S. Zhang, N.N. Sze, Real-time conflict risk at signalized intersection using drone video: A random parameters logit model with heterogeneity in means and variances, Accid. Anal. Prev. 207 (2024) 107739. https://doi.org/10.1016/j.aap.2024.107739

    work page doi:10.1016/j.aap.2024.107739 2024

  31. [31]

    Hossain, X

    A. Hossain, X. Sun, S. Das, M. Jafari, A. Rahman, Investigating pedestrian-vehicle crashes on interstate highways: Applying random parameter binary logit model with heterogeneity in means, Accid. Anal. Prev. 199 (2024) 107503. https://doi.org/10.1016/j.aap.2024.107503

    work page doi:10.1016/j.aap.2024.107503 2024

  32. [32]

    Y. Ali, F. Hussain, M.M. Haque, Advances, challenges, and future research needs in machine learning-based crash prediction models: A systematic review, Accid. Anal. Prev. 194 (2024) 107378. https://doi.org/10.1016/j.aap.2023.107378

    work page doi:10.1016/j.aap.2023.107378 2024

  33. [33]

    M. Yan, Y. Shen, Traffic Accident Severity Prediction Based on Random Forest, Sustainability 14 (2022). https://doi.org/10.3390/su14031729

    work page doi:10.3390/su14031729 2022

  34. [34]

    C. Chen, G. Zhang, Z. Qian, R.A. Tarefder, Z. Tian, Investigating driver injury severity patterns in rollover crashes using support vector machine models, Accid. Anal. Prev. 90 (2016) 128–139. https://doi.org/10.1016/j.aap.2016.02.011

    work page doi:10.1016/j.aap.2016.02.011 2016

  35. [35]

    Zheng, T

    M. Zheng, T. Li, R. Zhu, J. Chen, Z. Ma, M. Tang, Z. Cui, Z. Wang, Traffic Accident’s Severity Prediction: A Deep-Learning Approach-Based CNN Network, IEEE Access 7 (2019) 39897–39910. https://doi.org/10.1109/ACCESS.2019.2903319

    work page doi:10.1109/access.2019.2903319 2019

  36. [36]

    Niyogisubizo, L

    J. Niyogisubizo, L. Liao, Q. Sun, E. Nziyumva, Y. Wang, L. Luo, S. Lai, E. Murwanashyaka, Predicting Crash Injury Severity in Smart Cities: a Novel Computational Approach with Wide and Deep Learning Model, Int. J. Intell. Transp. Syst. Res. 21 (2023) 240–258. https://doi.org/10.1007/s13177-023-00351-7. 28

    work page doi:10.1007/s13177-023-00351-7 2023

  37. [37]

    Antariksa, R

    G. Antariksa, R. Tamakloe, J. Liu, S. Das, Automated and Explainable Artificial Intelligence to Enhance Prediction of Pedestrian Injury Severity, IEEE Trans. Intell. Transp. Syst. 26 (2025) 5568–

    2025

  38. [38]

    https://doi.org/10.1109/TITS.2025.3526217

    work page doi:10.1109/tits.2025.3526217 2025

  39. [39]

    Z. Wang, H. Guo, C. Zhang, Z. Hu, F. Zhou, Z. Sun, R. Sherony, S. Bao, Investigating pedestrian crash injury patterns: A comparative study of children and non-children, Accid. Anal. Prev. 222 (2025) 108223. https://doi.org/10.1016/j.aap.2025.108223

    work page doi:10.1016/j.aap.2025.108223 2025

  40. [40]

    M. Feng, J. Zhao, C. Hou, C. Nie, J. Hou, Investigating the safety influence path of right-turn configurations on vehicle–pedestrian conflict risk at signalized intersections, Accid. Anal. Prev. 211 (2025) 107910. https://doi.org/10.1016/j.aap.2024.107910

    work page doi:10.1016/j.aap.2024.107910 2025

  41. [41]

    Agheli, K

    A. Agheli, K. Aghabayk, How does distraction affect cyclists’ severe crashes? A hybrid CatBoost- SHAP and random parameters binary logit approach, Accid. Anal. Prev. 211 (2025) 107896. https://doi.org/10.1016/j.aap.2024.107896

    work page doi:10.1016/j.aap.2024.107896 2025

  42. [42]

    Goswamy, M

    A. Goswamy, M. Abdel-Aty, Z. Islam, Factors affecting injury severity at pedestrian crossing locations with Rectangular RAPID Flashing Beacons (RRFB) using XGBoost and random parameters discrete outcome models, Accid. Anal. Prev. 181 (2023) 106937. https://doi.org/10.1016/j.aap.2022.106937

    work page doi:10.1016/j.aap.2022.106937 2023

  43. [43]

    Z. Sun, D. Wang, X. Gu, M. Abdel-Aty, Y. Xing, J. Wang, H. Lu, Y. Chen, A hybrid approach of random forest and random parameters logit model of injury severity modeling of vulnerable road users involved crashes, Accid. Anal. Prev. 192 (2023) 107235. https://doi.org/10.1016/j.aap.2023.107235

    work page doi:10.1016/j.aap.2023.107235 2023

  44. [44]

    Scarano, M

    A. Scarano, M. Rella Riccardi, F. Mauriello, C. D’Agostino, N. Pasquino, A. Montella, Injury severity prediction of cyclist crashes using random forests and random parameters logit models, Accid. Anal. Prev. 192 (2023) 107275. https://doi.org/10.1016/j.aap.2023.107275

    work page doi:10.1016/j.aap.2023.107275 2023

  45. [45]

    Azmeri Khan, S

    S. Azmeri Khan, S. Yasmin, M. Mazharul Haque, Effects of design consistency measures and roadside hazard types on run-off-road crash severity: Application of random parameters hierarchical ordered probit model, Anal. Methods Accid. Res. 40 (2023) 100300. https://doi.org/10.1016/j.amar.2023.100300

    work page doi:10.1016/j.amar.2023.100300 2023

  46. [46]

    Sadeghi, K

    M. Sadeghi, K. Aghabayk, M. Quddus, A hybrid Machine learning and statistical modeling approach for analyzing the crash severity of mobility scooter users considering temporal instability, Accid. Anal. Prev. 206 (2024) 107696. https://doi.org/10.1016/j.aap.2024.107696

    work page doi:10.1016/j.aap.2024.107696 2024

  47. [47]

    Hossain, X

    A. Hossain, X. Sun, S. Das, M. Jafari, J. Codjoe, Investigating older driver crashes on high-speed roadway segments: a hybrid approach with extreme gradient boosting and random parameter model, Transp. Transp. Sci. 22 (2024) 1–35. https://doi.org/10.1080/23249935.2024.2362362

    work page doi:10.1080/23249935.2024.2362362 2024

  48. [48]

    CatBoost: gradient boosting with categorical features support

    A.V. Dorogush, V. Ershov, A. Gulin, CatBoost: gradient boosting with categorical features support, (2018). https://doi.org/10.48550/arXiv.1810.11363

    work page Pith review doi:10.48550/arxiv.1810.11363 2018

  49. [49]

    Prokhorenkova, G

    L. Prokhorenkova, G. Gusev, A. Vorobev, A.V. Dorogush, A. Gulin, CatBoost: unbiased boosting with categorical features, in: Adv. Neural Inf. Process. Syst., Curran Associates, Inc., 2018. https://proceedings.neurips.cc/paper_files/paper/2018/hash/14491b756b3a51daac41c24863285549- Abstract.html

    2018

  50. [50]

    Lundberg, S.-I

    S.M. Lundberg, S.-I. Lee, A unified approach to interpreting model predictions, in: Proc. 31st Int. Conf. Neural Inf. Process. Syst., Curran Associates Inc., Red Hook, NY, USA, 2017: pp. 4768– 4777

    2017

  51. [51]

    Andrea Cristina McGlinchey and Peter J

    S.M. Lundberg, G. Erion, H. Chen, A. DeGrave, J.M. Prutkin, B. Nair, R. Katz, J. Himmelfarb, N. Bansal, S.-I. Lee, From local explanations to global understanding with explainable AI for trees, Nat. Mach. Intell. 2 (2020) 56–67. https://doi.org/10.1038/s42256-019-0138-9

    work page doi:10.1038/s42256-019-0138-9 2020

  52. [52]

    Washington, M.G

    S. Washington, M.G. Karlaftis, F. Mannering, P. Anastasopoulos, Statistical and Econometric Methods for Transportation Data Analysis, 3rd ed., Chapman and Hall/CRC, New York, 2020. https://doi.org/10.1201/9780429244018

    work page doi:10.1201/9780429244018 2020

  53. [53]

    Hosmer, S

    D.W. Hosmer, S. Lemeshow, R.X. Sturdivant, Applied Logistic Regression, 3. Aufl, Wiley, Hoboken, N.J, 2013. 29

    2013

  54. [54]

    I. Mohamad, Quantifying the life-saving impact of seatbelt usage: A random forest analysis of unobserved heterogeneity and latent risk factors in vehicular fatalities, Multimodal Transp. 4 (2025) 100221. https://doi.org/10.1016/j.multra.2025.100221

    work page doi:10.1016/j.multra.2025.100221 2025

  55. [55]

    Fouda Mbarga, A.-R

    N. Fouda Mbarga, A.-R. Abubakari, L.N. Aminde, A.R. Morgan, Seatbelt use and risk of major injuries sustained by vehicle occupants during motor-vehicle crashes: a systematic review and meta- analysis of cohort studies, BMC Public Health 18 (2018) 1413. https://doi.org/10.1186/s12889-018- 6280-1

    work page doi:10.1186/s12889-018- 2018

  56. [56]

    Sarwahi, A.M

    V. Sarwahi, A.M. Atlas, J. Galina, A. Satin, T.J.I. Dowling, S. Hasan, T.D. Amaral, Y. Lo, N. Christopherson, J. Prince, Seatbelts Save Lives, and Spines, in Motor Vehicle Accidents: A Review of the National Trauma Data Bank in the Pediatric Population, Spine 46 (2021) 1637. https://doi.org/10.1097/BRS.0000000000004072

    work page doi:10.1097/brs.0000000000004072 2021

  57. [57]

    C. Lee, X. Li, Predicting Driver Injury Severity in Single-Vehicle and Two-Vehicle Crashes with Boosted Regression Trees, Transp. Res. Rec. 2514 (2015) 138–148. https://doi.org/10.3141/2514- 15

    work page doi:10.3141/2514- 2015

  58. [58]

    Islam, A.B

    S. Islam, A.B. Hossain, T.E. Barnett, Comprehensive Injury Severity Analysis of SUV and Pickup Truck Rollover Crashes: Alabama Case Study, Transp. Res. Rec. 2601 (2016) 1–9. https://doi.org/10.3141/2601-01

    work page doi:10.3141/2601-01 2016

  59. [59]

    Mannering, V

    F.L. Mannering, V. Shankar, C.R. Bhat, Unobserved heterogeneity and the statistical analysis of highway accident data, Anal. Methods Accid. Res. 11 (2016) 1–16. https://doi.org/10.1016/j.amar.2016.04.001

    work page doi:10.1016/j.amar.2016.04.001 2016

  60. [60]

    Fountas, P.Ch

    G. Fountas, P.Ch. Anastasopoulos, F.L. Mannering, Analysis of vehicle accident-injury severities: A comparison of segment- versus accident-based latent class ordered probit models with class- probability functions, Anal. Methods Accid. Res. 18 (2018) 15–32. https://doi.org/10.1016/j.amar.2018.03.003

    work page doi:10.1016/j.amar.2018.03.003 2018

  61. [61]

    Shannon, F

    D. Shannon, F. Murphy, M. Mullins, L. Rizzi, Exploring the role of delta-V in influencing occupant injury severities – A mediation analysis approach to motor vehicle collisions, Accid. Anal. Prev. 142 (2020) 105577. https://doi.org/10.1016/j.aap.2020.105577

    work page doi:10.1016/j.aap.2020.105577 2020

  62. [62]

    Arvin, A.J

    R. Arvin, A.J. Khattak, Driving impairments and duration of distractions: Assessing crash risk by harnessing microscopic naturalistic driving data, Accid. Anal. Prev. 146 (2020) 105733. https://doi.org/10.1016/j.aap.2020.105733

    work page doi:10.1016/j.aap.2020.105733 2020

  63. [63]

    Simmons, M

    S.M. Simmons, M. Donoghue, S. Erdelyi, H. Chan, C. Vaillancourt, P. Atkinson, F. Besserer, D.B. Clarke, P. Davis, R. Daoust, M. Émond, J. Eppler, J.S. Lee, A. MacPherson, K. Magee, E. Mercier, R. Ohle, M. Parsons, J. Rao, B.H. Rowe, J. Taylor, I. Wishart, J.R. Brubacher, Influence of cannabis and alcohol on motor vehicle injury severity in Canadian trauma...

    work page doi:10.1136/ip-2025-045642 2025

  64. [64]

    Islam, P

    M. Islam, P. Hosseini, A. Kakhani, M. Jalayer, D. Patel, Unveiling the risks of speeding behavior by investigating the dynamics of driver injury severity through advanced analytics, Sci. Rep. 14 (2024) 22431. https://doi.org/10.1038/s41598-024-73134-z

    work page doi:10.1038/s41598-024-73134-z 2024

  65. [65]

    Islam, A

    M. Islam, A. Mahmud, Unveiling the speeding behavior: Assessing the speeding risks and driver injury severities in single-heavy truck crashes, Saf. Sci. 187 (2025) 106861. https://doi.org/10.1016/j.ssci.2025.106861

    work page doi:10.1016/j.ssci.2025.106861 2025

  66. [66]

    Y. Chen, Y. Li, M. King, Q. Shi, C. Wang, P. Li, Identification methods of key contributing factors in crashes with high numbers of fatalities and injuries in China, Traffic Inj. Prev. 17 (2016) 878–

    2016

  67. [67]

    https://doi.org/10.1080/15389588.2016.1174774

    work page doi:10.1080/15389588.2016.1174774 2016

  68. [68]

    Hossain, X

    A. Hossain, X. Sun, S. Islam, S. Alam, Md. Mahmud Hossain, Identifying roadway departure crash patterns on rural two-lane highways under different lighting conditions: Association knowledge using data mining approach, J. Safety Res. 85 (2023) 52–65. https://doi.org/10.1016/j.jsr.2023.01.006

    work page doi:10.1016/j.jsr.2023.01.006 2023

  69. [69]

    Chakraborty, J

    R. Chakraborty, J. Liu, A.G. Tusti, M.S. Mimi, S. Das, Impact of lighting conditions on nighttime crash severity among older and elderly drivers, J. Transp. Saf. Secur. 17 (2025) 1377–1417. https://doi.org/10.1080/19439962.2025.2529833. 30

    work page doi:10.1080/19439962.2025.2529833 2025

  70. [70]

    Jafari Anarkooli, M

    A. Jafari Anarkooli, M. Hadji Hosseinlou, Analysis of the injury severity of crashes by considering different lighting conditions on two-lane rural roads, J. Safety Res. 56 (2016) 57–65. https://doi.org/10.1016/j.jsr.2015.12.003

    work page doi:10.1016/j.jsr.2015.12.003 2016

  71. [71]

    Roudsari, R

    B. Roudsari, R. Kaufman, R. Nirula, Comparison of mid-block and intersection-related left turn collisions, Traffic Inj. Prev. 8 (2007) 393–397. https://doi.org/10.1080/15389580701603227

    work page doi:10.1080/15389580701603227 2007

  72. [72]

    Asgarzadeh, S

    M. Asgarzadeh, S. Verma, R.A. Mekary, T.K. Courtney, D.C. Christiani, The role of intersection and street design on severity of bicycle-motor vehicle crashes, Inj. Prev. J. Int. Soc. Child Adolesc. Inj. Prev. 23 (2017) 179–185. https://doi.org/10.1136/injuryprev-2016-042045

    work page doi:10.1136/injuryprev-2016-042045 2017

  73. [73]

    Haque, H.C

    Md.M. Haque, H.C. Chin, H. Huang, Modeling fault among motorcyclists involved in crashes, Accid. Anal. Prev. 41 (2009) 327–335. https://doi.org/10.1016/j.aap.2008.12.010

    work page doi:10.1016/j.aap.2008.12.010 2009

  74. [74]

    Y. Li, J. Huang, Safety Impact of Pavement Conditions, Transp. Res. Rec. 2455 (2014) 77–88. https://doi.org/10.3141/2455-09

    work page doi:10.3141/2455-09 2014

  75. [75]

    Afghari, M.M

    A.P. Afghari, M.M. Haque, S. Washington, T. Smyth, Bayesian Latent Class Safety Performance Function for Identifying Motor Vehicle Crash Black Spots, Transp. Res. Rec. 2601 (2016) 90–98. https://doi.org/10.3141/2601-11

    work page doi:10.3141/2601-11 2016

  76. [76]

    Vertlib, S

    S.R. Vertlib, S. Rosenzweig, O.D. Rubin, A. Steren, Are car safety systems associated with more speeding violations? Evidence from police records in Israel, PLOS ONE 18 (2023) e0286622. https://doi.org/10.1371/journal.pone.0286622

    work page doi:10.1371/journal.pone.0286622 2023

  77. [77]

    J. Liu, J. Li, K. Wang, J. Zhao, H. Cong, P. He, Exploring factors affecting the severity of night- time vehicle accidents under low illumination conditions, Adv. Mech. Eng. 11 (2019) 1687814019840940. https://doi.org/10.1177/1687814019840940

    work page doi:10.1177/1687814019840940 2019

  78. [78]

    Zhang, M

    K. Zhang, M. Hassan, Crash severity analysis of nighttime and daytime highway work zone crashes, PLOS ONE 14 (2019) e0221128. https://doi.org/10.1371/journal.pone.0221128

    work page doi:10.1371/journal.pone.0221128 2019

  79. [79]

    C. Lyon, B. Persaud, D. Merritt, J. Cheung, Empirical Bayes Before-After Study to Develop Crash Modification Factors and Functions for High Friction Surface Treatments on Curves and Ramps, Transp. Res. Rec. 2674 (2020) 505–514. https://doi.org/10.1177/0361198120957327

    work page doi:10.1177/0361198120957327 2020

  80. [80]

    Cheng, R

    G. Cheng, R. Cheng, Y. Pei, J. Han, Research on Highway Roadside Safety, J. Adv. Transp. 2021 (2021) 6622360. https://doi.org/10.1155/2021/6622360

    work page doi:10.1155/2021/6622360 2021

Showing first 80 references.