In the past few years, Road accidents have been are increasing in most of the developing countries, including India. The main causes of the road accidents are drunken driving, improper vehicle condition, and rash driving. Traffic collision may result in injury, death, vehicle damage, and property damage. The analysis and identification of such accident prone zones are essential to reduce the same. The aim of this paper is to identify the crash prone zones, and to develop regression models using multiple linear regression and negative binomial models, which predict the crashes. Crash data is collected for different intersections of the selected segments. Data was collected for the year 2011 to 2015, from Hyderabad police stations in terms of crashes per year, traffic volume (major and minor road), turning traffic volume in vehicle per day etc. These data are analysed, and a regression model is developed using various intersection parameters. The model was developed to show the relationship between crash rate and various intersection parameters from regression analysis. If the number of intersections and vehicles increase, then crashes are going to increase. Hence, the model has been developed to reduce the collision.