Predicting Freeway Crash Likelihood and Severity with Real-time Loop Detector Data

摘要

Real-time crash risk prediction using traffic data collected from loop detector stations is useful indynamic safety management systems aimed at improving traffic safety through application ofproactive safety countermeasures. The major drawback of most of the existing studies is thatthey focus on the crash risk without consideration of crash severity. This paper presents an effortto develop a model that predicts the crash likelihood at different levels of severity with aparticular focus on severe crashes. The crash data and traffic data used in this study werecollected on the I-880 freeway in California, United States. This study considers three levels ofcrash severity: fatal/incapacitating injury crashes (KA), non-incapacitating/possible injurycrashes (BC), and property-damage-only crashes (PDO). The sequential logit models weredeveloped to link the likelihood of crash occurrences at different severity levels to various trafficflow characteristics derived from detector data. The fitness and prediction capability of theforward and backward versions of the models were compared to select a better alternative.The results show that the sequential structure (forward vs. backward) does not have considerableimpact on the model's fitness and predictive capabilities. More interestingly, the traffic flowcharacteristics contributing to crash likelihood were quite different at different levels of severity.The PDO crashes were more likely to occur under congested traffic flow conditions with highlyvariable speed and frequent lane changes, while the KA and BC crashes were more likely tooccur under less congested traffic flow conditions. High speed, coupled with a large speeddifference between adjacent lanes under uncongested traffic conditions, was found to increasethe likelihood of severe crashes (KA). This study applied the 20-fold cross-validation method toestimate the prediction performance of the developed models. The validation results show thatthe model's crash prediction performance at each severity level was satisfactory. The findings ofthis study can be used to predict the probabilities of crash at different severity levels, which isvaluable knowledge in the pursuit of reducing the risk of severe crashes through the use ofdynamic safety management systems on freeways.