Alternative Method for Identifying Crash Hotspot Using Detailed Crash Information from First Information Report (FIR)

Identification and improvement of crash hotspots is an important step towards road safety improvement. The Hotspot Identification (HSID) methods aim to identify a list of hotspots with minimum false identifications. HSID methods available in literature identifies hotspots using crash history information, traffic information and road geometry information. The crash history-based methods use Crash Frequency (CF) and severity (Equivalent Property Damage Only—EPDO) information only. The crash time and weather information, which determine the visibility and pavement conditions, are not considered. Researchers claim that simultaneous use of multiple HSID methods yield minimum false identifications. This work aims to utilize weather and time information, by introducing a Cause Index (CI) and a Time Index (TI), respectively, along with conventional EPDO method to systematically identify a group of hotspots using a multi-dimensional K-means clustering algorithm. The causes assigned in the CI are skid, night, fog and faulty geometry. TI categorizes crashes into day-time and night-time crashes. Use of only one or simultaneous multiple HSID measures to designate hotspots is not sufficient as these are based on three-year average crash occurrences. The crash occurrence pattern needs to be studied. The monthly crash occurrence pattern, checked using Runs Test, can confirm the hotspot list. In the proposed method, number of candidate hotspots is not a function of practitioners’ discretion. The proposed method was tested with road crash data of Patna, Bihar, India and its performance was compared with CF and EPDO methods. It was observed that the proposed method yielded minimum false identifications.