Modeling accident frequency in Denmark for improving road safety

Allan Lyckegaard, Tove Hels, Sigal Kaplan, Inger Marie Bernhoft, Carlo Giacomo Prato

    Research output: Contribution to conferenceConference abstract for conferenceResearch

    2 Downloads (Pure)


    Traffic accidents result in huge costs to society in terms of death, injury, lost productivity, and property damage. The main objective of the current study is the development of an accident frequency model that predicts the expected number of accidents on a given road segment, provided the infrastructure characteristics and the traffic conditions of the road. The model can be used to point out high risk road segments and support road authorities in planning interventions for the improvement of road safety on Danish roads.
    The number of accidents on a road link was modeled using a count model after verifying the presence of overdispersion in the variance of the counts. The model relates the number of accidents to the characteristics of the infrastructure in terms of geometry and traffic in order to identify which risk factors and at which degree relate to the possibility of accident occurrence. Data concerning police recorded accidents, link characteristics of the road network, traffic volumes from the national transport models are merged to estimate the model. Spatial correlation between road sections is taken into account for correcting for unobserved correlation between contiguous locations.
    Original languageEnglish
    Publication date2013
    Number of pages1
    Publication statusPublished - 2013
    EventStrategisk forskning i transport og infrastruktur - Danmarks Tekniske Universitet, Kongens Lyngby, Denmark
    Duration: 11 Jun 201312 Jun 2013


    ConferenceStrategisk forskning i transport og infrastruktur
    LocationDanmarks Tekniske Universitet
    CityKongens Lyngby
    Internet address


    Dive into the research topics of 'Modeling accident frequency in Denmark for improving road safety'. Together they form a unique fingerprint.

    Cite this