Planning Level Regression Models for Prediction of the Number of Crashes on Urban Arterials in Bangladesh

In most of the developing countries, the metropolitan organizations do not assess the safety consequences of alterna- tive transportation systems and one of the reasons is the lack of suitable methodology. The goal of this paper is to develop practical tools for assessing safety consequences of arterial roads in the context of long-term urban transpor- tation plans in Dhaka city, the capital of Bangladesh. The researchers used the generalized linear modeling approach to develop separate models to predict number of crashes for different levels of crash severity for major arterial seg- ments which have the highest crash rates. The models used five independent variables - length of segment, traffic volume, number of access road, design speed and roadway width - all of which are usually collected or predicted by transportation planners. The study reveals that roadway width and design speed are the governing factors for non-lane based traffic on urban arterials for controlling crashes. The crash prediction models presented in this paper can enable Dhaka’s transport planners to evaluate the safety impact of alternative road networks with regard to the costs and benefits in long-term planning context.

Keywords:

Crash prediction , regression models , negative-binomial distribution , poisson distribution , fatal accident