Development of Models for Crash Prediction and Collision Estimation- A Case Study for Hyderabad City

Road traffic crash is a cause of unnatural death and occupies fifth position in the world as per WHO records. Road crashes in India are alarming in situation while road safety is professionally lacking and politically missing. Hyderabad city, the capital of newly formed Telangana State occupies sixth position in occurrence of road crashes. An attempt is made to understand which model is suitable for road crash prediction and estimation of collision type which is influencing road crashes in the city. A retrospective observational study was conducted in the city of considering factors as roadway geometrics, traffic data and type of collision. Regression models like Multiple Linear Regression, Poisson Regression, Logit Model and Multinomial Logistic Model are considered for arriving crash prediction models. The suitability of model is based on R2 and chi-square test. It is observed from the analysis through comparison of above model; Logit model has an R2 of 0.7 and is significant for chi-square test. Shoulder condition is an affecting factor for non-fatal crashes in the city which is ascertained by Logit model. Further analysis is also carried for arriving on type of collision influencing on road crashes. It is observed from the results that sideswipe is more significant for road crashes.