Incident duration model and secondary incident causation model based on archived traffic management center data

Abstract

Highway incidents are a major source of traffic congestion, the paramount operational problem on urban freeways. Considerable national attention has been directed towards the effective management of traffic incidents to alleviate congestion-related problems on freeways. Evaluating such incident management programs requires developing models that can characterize incident properties. This study provides two statistical models for predicting incident durations and secondary incident causation probabilities based on eleven months of incident data from the Nashville Traffic Management Center. The non-linear relationship between the predictor variables (descriptive incident characteristics) and the response variables (incident duration and secondary incident causation probability) was accounted for by using logistic regression methods. In the secondary incident causation model the clearance time of an incident was used as the sole explanatory variable to predict the secondary incident causation probabilities.