Spatial Analysis of Pedestrian Accidents in District 5 of Tehran

This study presents an approach that uses kernel density function algorithm and spatial autocorrelation analysis in an integrated way to facilitate the identification of accident prone areas and simultaneously evaluate the statistical significance of dense clusters. For this purpose, at first, data related to pedestrian accidents from 1995 to 1999 in District 5 of Tehran city were extracted from accident maps and their geographic coordinates were recorded in the GIS environment. Incident points were identified using KDE algorithm based on geographic information system. Then, hot clusters were evaluated for statistical significance using nearest neighbor, global Moran's and general G statistics. Finally, Anselin's local Moran's index was used to show the statistical distribution of phenomena in space and to analyze clusters and non-clusters. In general, the results of spatial statistics indicate strong clustering and the formation of clusters with a high density of accidents next to each other.