Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM68531
Volume:
10 |
10 |
Oct 2026
Pages:
1-4
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
Information Technology
Shah & Anchor Kutchhi College of Engineering
Department of Information Technology
Shah & Anchor Kutchhi College of Engineering
- Road accidents are a global issue however India has recorded most fatalities which accounts for roughly 11% of all road-related deaths worldwide. Over 53% of all accidents and 59% of all fatalities occurred on highways. This study proposes a method to classify a road accident as Fatal, Grievous, Minor or No-Injury. The dataset used for this study comprises of road accident data collected from four distinct highway segments under the jurisdiction of the National Highways Authority of India (NHAI). To handle class imbalance, SMOTE with Tomek links were used. Various machine learning models were used such as Random Forest, XGBoost, LightGBM. The models achieved accuracies of 86.2%, 84.7%, and 83.4% respectively. As most of the machine learning models exhibit a common limitation: “black box” nature, results were interpreted using SHapley Additive exPlanations (SHAP) method. SHAP analysis reveals that accident location represented by highway chainage and the primary vehicle type consistently emerged as the dominant predictors for both Grievous and Fatal accident classifications. Moreover, accident cause exhibited greater prominence in the Fatal-class explanation, whereas temporal characteristics occupied higher ranks in the Grievous-class explanation.
KEYWORDS
This study presented a hybrid explainable machine learning framework for predicting road accident severity by utilizing the strengths of ensemble and boosting models with interpretability techniques. The results demonstrated that Random Forest model is highly effective in capturing complex patterns within accident data, while explainable AI approaches provide valuable insights into how different factors influence severity outcomes. By moving beyond simple prediction and offering interpretable results, the study contributes towards more transparent and reliable decision-support systems for road safety. Looking ahead, several avenues can further enhance this work. Future research can focus on incorporating real-time data streams instead of relying solely on static or historical datasets, enabling dynamic and more responsive prediction systems. In addition, counterfactual explainable AI methods, such as those implemented through tools like DiCE, can be explored to provide actionable “what-if” scenarios, helping stakeholders understand how changes in specific factors could alter accident outcomes.
[1] H. Rosen, I. Bari, and N. Paichadze, “Global road safety 2010–2018: An analysis of Global Status Reports,” Injury, vol. 56, 2022. [2] “ROAD ACCIDENTS IN INDIA 2022.”, Highways Accidents Report, Ministry of Road Transport and Highways, 2022. www.morth.nic.in. [3] Santos, K., et al. "A literature review of machine learning algorithms for crash injury severity prediction.." Journal of safety research, vol. 80, 2021, pp. 254-269, https://doi.org/10.1016/j.jsr.2021.12.007. [4] Ahmed, Shakil, et al. "A study on road accident prediction and contributing factors using explainable machine learning models: analysis and performance." Transportation Research Interdisciplinary Perspectives, 2023, https://doi.org/10.1016/j.trip.2023.100814. [5] Khanum, Humera, et al. "Accident severity prediction modeling for road safety using random forest algorithm: an analysis of Indian highways." F1000Research, vol. 12, 2023, https://doi.org/10.12688/f1000research.133594.2. [6] Khanum, Humera et al. “A methodological framework for road accident severity prediction for indian highways using machine learning models.” MethodsX, vol. 15, pp. 103728, 19 Nov. 2025, doi:10.1016/j.mex.2025.103728.
Publication Date
10/31/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
2
Total views since publication
PDF Downloads
3
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...