IS-BASED ROAD ACCIDENT ANALYSIS AND KERNEL DENSITY ESTIMATION: A REVIEW OF HOTSPOT IDENTIFICATION METHODS, APPLICATIONS, AND RESEARCH GAPS

Road traffic accidents are a major public - safety concern, and identifying accident-prone locations is essential for effective road-safety planning. Geographic Information Systems (GIS) provide an integrated platform for mapping accident locations, analysing spatial patterns, and supporting the prioritisation of road-safety interventions. Kernel Density Estimation (KDE) is widely used to transform individual accident locations into continuous density surfaces, enabling the visual identification of areas with high concentrations of crashes. This review examines the applications of GIS and KDE in road accident analysis, with particular attention to conventional KDE, network KDE, hotspot validation, and integration with statistical and machine-learning techniques. Anderson (2009) demonstrated the use of GIS-based KDE and K-means clustering to identify and characterise injury-accident hotspots in London. Xie and Yan (2008) developed a network-based KDE approach to account for the fact that traffic accidents occur along road networks. Thakali et al. (2015) compared KDE and kriging for crash hotspot identification, while Harirforoush and Bellalite (2019) combined network KDE with crash-rate screening to improve hotspot assessment. These studies show that KDE is useful for exploratory spatial analysis, but a high density of recorded crashes does not necessarily establish a statistically significant hotspot or a high exposure-adjusted risk.