INCIDENT-DETECTION ALGORITHMS. PART 2. ON-LINE EVALUATION

Five algorithms were evaluated on-line by using the facilities of the Traffic Systems Center of the Illinois Department of Transportation. Three of the algorithms developed by Technology Services Corporation (TSC), were of a pattern-recognition nature. The other two--a pattern-recognition and a probabilistic or Bayesian algorithm--were developed locally. Thresholds for the features used in each of the pattern-recognition algorithms were developed by TSC. The thresholds for the probabilistic algorithm were developed by using accident data on the Eisenhower Expressway. The measures of effectiveness in the evaluation were detection rate, false-alarm rate, and mean-time-to-detect. The three TSC algorithms were evaluated twice on the Eisenhower Expressway at the 80 and 90 percent levels of detection thresholds, and then problem areas showing high false-alarm rates were represented by the 50 percent level. The three TSC algorithms were then evaluated on a section of the Dan Ryan Expressway that was free of geometric problems, for comparison purposes. Statistical analysis showed no difference in detection rate, false-alarm rate, and mean-time-to-detect among the three TSC algorithms at any of the evaluated detection levels. Introduction of the 50 percent level improved certain measures of effectiveness. Algorithm 7, the best of the TC algorithms, showed overall superiority to the two local algorithms. The false-alarm rate was shown to be related to geometric and other features of the problem areas and yielded algorithm 8, which uses a shock wave-suppressor mechanism and requires the least effort in developing appropriate thresholds. /Author/

Media Info

  • Media Type: Print
  • Features: Figures; Maps; Tables;
  • Pagination: pp 58-64
  • Monograph Title: Urban systems operations
  • Serial:

Subject/Index Terms

Filing Info

  • Accession Number: 00308582
  • Record Type: Publication
  • ISBN: 0309029724
  • Files: TRIS, TRB
  • Created Date: Apr 22 1980 12:00AM