Density-based clustering for road accident data analysis

Alotaibi, Abdullah Shawan · 2018 · OpenAlex-citations

DOI: 10.21833/ijaas.2018.08.014

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 critical issue of road traffic accidents, which cause significant loss of life and injury, by employing data mining techniques to identify underlying factors and patterns. The research is motivated by the need to move beyond traditional statistical methods, which often struggle with the heterogeneity and high dimensionality of accident data. The primary objective is to recognize key issues in road safety by discovering associations between various attributes—such as human, vehicle, and infrastructure factors—to enable predictive analysis and effective preventive measures. The methodology utilizes a two-stage data mining approach implemented using the Weka tool. First, the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm is applied to partition accident locations into clusters based on accident frequency. DBSCAN is selected for its ability to handle large datasets with noise and determine clusters of varying shapes without requiring a predefined number of clusters. The algorithm classifies data points as core, boundary, or noise based on density thresholds defined by parameters $\epsilon$ (radius) and MinPts (minimum points). Second, the Parallel Frequent Pattern Mining algorithm, specifically FP-growth, is applied to these clusters. This association rule mining technique extracts hidden relationships between different attributes within each cluster to reveal specific features and contributing factors for accidents in those locations. The analysis was conducted on a road accident dataset comprising 11,574 records from 2012 to 2016. After preprocessing to remove noise, 11 variables were identified for analysis, including accident time, type, number of injured victims, age and gender of victims, road type, area, number of vehicles involved, road surface, lighting conditions, weather conditions, and geographic coordinates (Easting and Northing). The study also contextualizes the problem with statistical data from India, noting that driver faults account for 79% of accidents, with speeding being the largest contributor at 55.6%, while environmental and road design factors account for less than 5% combined. The results demonstrate that the combination of DBSCAN and parallel frequent mining effectively explores accident data to uncover patterns and predict future attitudes. By clustering locations and mining frequent itemsets, the approach identifies specific associations between attributes that contribute to accidents in different areas. The study concludes that this data mining framework allows for the efficient identification of key issues in road safety, providing a basis for targeted preventive measures. The ability to detect hidden patterns and predict accident-prone conditions supports transportation agencies in implementing strategies to reduce accidents and improve road safety.

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.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• telematics crash prediction

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