A data mining approach for road accident analysis with safety planning using K-means clustering and association rules
Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering · 07.05.2026
DOI: 10.1177/09544070261446547 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, online erschienen am 07.05.2026. Verfasst von Santhiya Ravindran und 2 weiteren Autor:innen. Die Arbeit nennt 63 Literaturangaben.
Abstract (Originalsprache)
Road traffic accidents (RTAs) represent a major public safety concern worldwide and identifying influential crash risk factors is essential for developing effective primary safety and prevention strategies. Although numerous studies have applied traditional data mining techniques to analyze RTA data, the potential of hybrid modeling approaches remains insufficiently explored. To address this gap, the present study proposes a comprehensive multi-method analytical framework that integrates statistical visualization techniques with hybrid data mining methods to gain deeper insights into road crash patterns. Initially, histogram plots with overlaid normal curves are employed to examine the distribution characteristics of key accident related variables. Multiple Correspondence Analysis (MCA) is then applied to visualize and interpret complex interrelationships among categorical crash attributes. Subsequently, K-means clustering based hybrid association rule mining (ARM) models (Hybrid Apriori, ECLAT, FP-Growth with all features and with selected features) are implemented using accident data collected from 49 regions of Virudhunagar district, Tamil Nadu, India. The performance of these hybrid models is systematically compared with classical ARM algorithms (Classical Apriori, ECLAT, and FP-Growth with all features) by analyzing 2-cause, 3-cause, and 4-cause associated frequent patterns. The results indicate that two-wheelers are the most frequently involved vehicles, with hit-from-back and head-on collisions emerging as the dominant crash types, particularly during night time. Pedestrians are identified as highly vulnerable road users. Compared with classical approaches, hybrid ARM models extract richer and more meaningful crash risk patterns. Notably, hybrid models with selected features demonstrate superior computational efficiency, achieving execution times as low as 156 ms and memory usage of 4.59 MB.
Zitieren
APA
Ravindran, S., Balachandran, G. B., & Velladurai, P. (2026). A data mining approach for road accident analysis with safety planning using K-means clustering and association rules. Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering. https://doi.org/10.1177/09544070261446547
Vancouver
Ravindran S, Balachandran GB, Velladurai P. A data mining approach for road accident analysis with safety planning using K-means clustering and association rules. Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering. 2026. doi:10.1177/09544070261446547
BibTeX
@article{Ravindran2026,
author = {Ravindran, Santhiya and Balachandran, Gurukarthik Babu and Velladurai, Petchithai},
title = {A data mining approach for road accident analysis with safety planning using K-means clustering and association rules},
journal = {Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering},
year = {2026},
doi = {10.1177/09544070261446547}
}
EUVIMED