Black spot identification on two lane rural roads using accident prediction model

Accident prediction models (APM) are an important instrument for evaluating road safety performance. These models explain accident occurrence as a function of traffic and geometric characteristics of road. This paper previously describes the development of accident prediction model (APM) for two lane rural roads and then illustrates an application to identify hazardous sites in a road network. The generalized linear model approach (GLIM) is used to calibrate two different APMs, one using as explanatory variable both the traffic and the length of road segments and the other one with only one variable to explain the exposure. The procedure based on GLIM has the advantage of overcoming the limitations due to conventional linear regression in accident frequency modelling. In order to identify black spots both linear and non linear model were used and the Empirical Bayes technique was applied with the aim to smooth out the random fluctuation in accident data. Finally, the comparison between linear and non linear models shows that the former can produce incorrect black spot identification (false negative), especially on sites showing low level of exposure (A). For the covering abstract of the conference see ITRD E212343.

  • Authors:
    • CAFISO, S
    • AUGERI, M G
    • LA CAVA, G
  • Publication Date: 2006


  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01102569
  • Record Type: Publication
  • Source Agency: TRL
  • Files: ITRD
  • Created Date: Jun 16 2008 7:59AM