The incident identification process uses an incident detection algorithm to relate certain measured relationships between traffic characteristics to calibrated ones giving a prediction regarding the occurrence of an incident. The development of the algorithm depends upon pattern recognition and statistical forecasting of traffic behaviour. The efficiency of such algorithms can be determined by the detection rate, the false alarm rate and the mean time taken to detect an incident. Bayesian concepts are used to incorporate incident historical information into a probabilistic model. The model is not limited to the two possible messages of incident occurred or incident free, but allows estimation of the likelihood of an incident signal being a false alarm. It would be possible to have a string of feature values displayed to enable an operator to judge the severity of the situation. The efficiency of the algorithm was evaluated by running through incident and incident-free data related to the study site. The results were also compared with those from the applications of three other algorithms. /TRRL/

  • Availability:
  • Corporate Authors:

    Printerhall Limited

    29 Newmart Street
    London W1P 3PE,   England 
  • Authors:
    • Levin, M
    • Krause, G M
  • Publication Date: 1979-3

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 00302565
  • Record Type: Publication
  • Source Agency: Transport Research Laboratory
  • Files: ITRD, TRIS
  • Created Date: Jan 30 1980 12:00AM