In the context of pro-active traffic management, real-time crash risk evaluation is one of the most important components. Signalized intersections are well-known high-risk locations because of the variety of traffic movements, modes, and their interactions. Unlike access-controlled freeways, the traffic flow at signalized intersections presents cyclical characteristics, which are temporally interrupted by signal timing. Therefore, the data preparation for real-time crash risk prediction at signalized intersections should be based on the signal cycle rather than a predefined fixed time interval (i.e., 5 minutes). In this research, the actual signal cycles where crashes have occurred were identified based on high-resolution event-based data (i.e., Automated Traffic Signal Performance Measures (ATSPM)). Meanwhile, all the real-time cycle-level data were calculated, including traffic volume, signal timing, headway and occupancy, traffic variation between upstream and downstream detectors, shockwave characteristics, and weather data. In this study, two approaches of undersampling strategies (i.e., matched case-control and random sampling) were utilized to develop conditional logistic and binary logistic models, respectively. Model results indicated that the binary logistic model based on the random undersampled dataset performs much better than the conditional logistic model based on the matched case-control dataset. It is revealed that higher cycle volume, overall average flow ratio across lanes, arrivals on yellow ratio, traffic volatility across approach sections, as well as longer cycle length and lower green ratio could significantly increase the crash likelihood at signalized intersections. Moreover, longer queue length, bigger shockwave, and higher absolute queuing shockwave speed tend to increase the crash likelihood.

CICTP