Hot-spot Identification: a Categorical Binary Model Approach

Abstract

This paper presents an alternative methodology for hot-spot identification based on a probabilistic model. In this methodology, the ranking criterion for hot-spot identification conveys the probability of a site being a hot-spot or a non-hot spot. A binary choice model was used to link the outcome to a set of factors that characterize the risk of the sites under analysis based on our use of two categories (0/1) for the dependent variable. The proposed methodology consists of two main steps. First, a threshold value for the number of accidents is set to distinguish hot spots from safe sites (category 1 or 0, respectively). Based on this classification, a binary model is applied that allows the construction of an ordered site list using the probability of a site being a hot-spot. The second step involves the choice of a selection strategy. The selection strategy can target a fixed number of sites with the greatest probability or, alternatively, all sites exceeding a specific probability, such as 0.5. A demonstration of the proposed methodology is provided using simulated data. For the simulation design, urban intersection data from Porto, Portugal, covering a five-year period were used. The results of the binary model showed a good fit. To evaluate and compare the probabilistic method with other commonly used methods, measures were used to test the performance of each method in terms of its power to detect the "true" hot spots. The test results indicate that the proposed method is superior to two commonly used methods. The gains of using this method are related to the simplicity of its application, while critical issues such as prior distribution effect assumptions and the regression-to-the-mean phenomenon are overcome. Further, the proposed model provides a realistic and intuitive perspective and supports easy practical application.