RARE SEVERE PATTERNS IN ROAD ACCIDENT RECORDS: A REVIEW

Authors

  • Desai Saurabh Sudhir Dr. A.P.J. Abdul Kalam University, Indore, M.P., India Author
  • Arpit Solanki Dr. A.P.J. Abdul Kalam University, Indore, M.P., India Author

Keywords:

Association rule mining; frequent pattern mining; FP-Growth; class association rules; road traffic accidents; crash severity; interestingness measures; false discovery rate.

Abstract

Road traffic crashes remain a major public-health burden, and the police records that describe them are a rich but under-used source of safety knowledge. Association rule mining turns such records into readable if–then statements, which makes it attractive for road-safety work. This paper reviews the literature that bears on applying it to crash data: the foundations of frequent pattern mining and the efficiency techniques built on them, condensed representations, multiple-support and weighted mining, class association rules, interestingness measures, class imbalance, the statistical significance of discovered patterns, and data mining studies of road accidents. The surveyed methods are then compared in terms of what each contributes and what limits it for severity-labelled crash records, with attention to weighting, class-specific support thresholds, class-targeted mining and removal of redundant rules. The comparison shows that each property has been studied in isolation, mostly on generic benchmark data, whereas road-safety applications generally use existing miners without changing them. No reviewed method combines severity-aware weighting, class-specific thresholds, class-targeted search, redundancy control and protection against chance rules, and this absence is identified as the research gap. Based on this gap, the paper points to one possible direction for future work.

Downloads

Published

2026-11-16