Estimating freeway incident duration using accelerated failure time modeling

Effective incident management requires a good understanding of various characteristics of incidents in order to accurately estimate incident durations and help make more efficient decisions to reduce the impact of the incident. The objective of this study is to identify the factors affecting the incident duration. AIC (Akaike Information Criterion) is used to find that the group of reporting period, time for the policeman to reach the site, incident type, lanes blocked, vehicle type, vehicles involved, deaths and number of rescue vehicles is the best for freeway incident duration prediction. Accelerated Failure Time (AFT) model is applied to estimate the incident duration and the model tests indicate that the log-logistic distribution produces the best fit between the incident duration and Chinese policeman’s freeway incident record. In addition using Accelerated Failure Time (AFT) model the common incident duration modeling difficulty of missing data is solved. The results show that policeman’s reaching site time is the detrimental factor on incident duration, and if the policeman takes less than 30 min to arrive the site, the incident would last much less time. Incident in the night takes longer time. The fire incident which affects many government departments or rescue agencies takes significantly longer time. Besides the fire, rollover accident which need more clearance time lasts longer than any other incident. Severe accidents relating to more deaths and rescuer vehicles often take more incident time.

► Accelerated Failure Time (AFT) model is employed to estimate the freeway incident duration.
► Log-logistic distribution produces the best fit for the freeway incident duration.
► Policeman’s reaching site time is the detrimental factor on incident duration.
► The fire incident takes significantly longer time.