Calibrating a Real-Time Traffic Crash-Prediction Model Using Archived Weather and ITS Traffic Data

Growing concern over traffic safety has led to research efforts directed towards predicting freeway crashes in Advanced Traffic Management and Information Systems (ATMIS) environment. This paper aims at developing a crash-likelihood prediction model using real-time traffic-flow variables (measured through series of underground sensors) and rain data (collected at weather stations) potentially associated with crash occurrence. Archived loop detector and rain data and historical crash data have been used to calibrate the model. This model can be implemented using an online loop and rain data to identify high crash potential in real-time. Principal component analysis (PCA) and logistic regression (LR) have been used to estimate a weather model that determines a rain index based on the rain readings at the weather station in the proximity of the freeway. A matched case-control logit model has also been used to model the crash potential based on traffic loop data and the rain index. The 5-min average occupancy and standard deviation of volume observed at the downstream station, and the 5-min coefficient of variation in speed at the station closest to the crash, all during 5-10 min prior to the crash occurrence along with the rain index have been found to affect the crash occurrence most significantly.