Integration of k-Means Clustering and Association Rule Mining for Road Safety Analysis

Abstract

Data mining has proven to be a reliable technique for analyzing road accidents and yielding productive results. Many analyses of road traffic accident data employ data mining techniques, with a focus on identifying factors influencing accident severity. Nonetheless, any damage resulting from road accidents is consistently deemed unacceptable due to its implications on health, property, and various economic factors. Occasionally, it is observed that road accidents occur more frequently and pose higher risks at specific locations. Analyzing these locations can aid in identifying key features contributing to the frequent occurrence and increased risk of accidents.One popular data mining technique for identifying correlations in various attributes of road traffic accidents is association rule mining. In this paper, The association rules were generated through the application of the Frequent Pattern (FP) Growth Algorithm. To facilitate this analysis, the k-Means algorithm was utilized to categorize locations into three groups: High-Risk Hotspots (HRH), Moderate-Risk Zones (MRZ), and Low-Risk Areas (LRA) based on their Accident Frequency Count and Accident Average Severity Score. This approach focused on locations with diverse risk levels, facilitating an analysis of frequent incidents, rather than concentrating solely on areas with a high frequency of accidents. The generated rules unveiled different factors associated with road accidents at locations exhibiting varying Accident Frequency Count and Accident Average Severity Score. Our approach uncovered some valuable hidden insights from the data which can be utilized to make some preventive efforts in these locations.