A Hybrid Logistic Regression and Spatial Analysis Approach for Identification of Candidate Highway Safety Projects

Highway safety continues to be one of the most critical issues facing highway agencies. This paper proposes a hybrid approach for the identification of safety projects via a combination of a conventional regression-based model and spatial analysis techniques. For many transportation agencies, safety projects are identified and prioritized via cost-benefit analyses, multi-attribute ranking criteria, crash frequency modeling, statistical models, and many others. Nonetheless, for many conventional statistical models in the existing literature, the studies only provide statistically significant explanatory variables which influence the frequency and/or injury severity of crashes. More often than not, there is no specific approach, based on the statistical model, which guides highway safety engineers in the identification of high-risk road segments for targeted interventions. Furthermore, there is a deficiency in formalized frameworks which utilizes generalized linear models in tandem with spatial tools to propose specific safety projects based on crash severity. The proposed methodology leverages the generality of findings using regression-based methods with spatial querying and a hot spot analysis (using the Getis-Ord Gi* statistic) method to identify potential locations for safety projects. A synthesized framework is proposed and implemented using crash data for Tarrant County in Texas. This approach allows the integration of expert knowledge from safety engineers and also provides a pragmatic approach to ameliorating crash injury severity on the highway network.