Development of Model for Road Crashes and Identification of Accident Spots

Abstract

The aim of this study is to assess the safety of multi-lane rural highway in India. This paper shows the application of a generalized linear modeling technique for the analysis of road accidents on the Indian National Highway. Speed, traffic flow and road characteristics data on four-lane dived rural highway in Dahod are analyzed. The study proposes a novel approach to include average daily traffic (ADT) and average spot speed (AS) in the accident prediction model for a rural highway. The model has been developed for accidents per km as a dependent variable and significant variables such as Junction density, village settlement nearby, ADT, AS as independent variables. The findings from the model offer a better estimate of accidents for a multilane divided rural highway. Statistical Models cannot fully reflect the characteristics of each section due to the heterogeneous nature of road accidents, so the association rule mining technique has been used to identify accident spots as it can deal with the heterogeneous nature of accidents. Accident spots have been assessed by correlating various attributes to the severity of the accident (fatal, non-fatal). This research will help to improve road safety on rural highways.