Experimental Evaluation of Hotspot Identification Methods

The first part of this paper compares alternative hotspot identification (HSID) methods. First, crash frequency instead of crash rate data are used to assess hotspot identification techniques; second, the paper compares the performance of simple ranking, classical confidence intervals, and empirical Bayesian techniques in terms of percent false negatives and positives; third, several practical crash distributions from the state of Arizona are selected to represent a realistic range of 'base' crash data; and finally, several degrees of crash heterogeneity are examined in the experimental evaluation. Three HSID methods observed in practice are evaluated. The paper also evaluates the effect of crash history duration employed in the three HSID methods. The examination of the simulated results illustrates that the Empirical Bayes technique significantly outperforms ranking and confidence interval techniques; false positives and negatives are inversely related; and three years of crash history appears to provide an appropriate crash history duration.

Language

  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01004171
  • Record Type: Publication
  • Files: TRIS, ATRI
  • Created Date: Sep 16 2005 10:51AM