High-order Markov model for prediction of secondary crash likelihood considering incident duration.

Traffic incidents can create major non-recurring congestion and have the potential to be fatal. Traffic Engineers and researchers have worked vigorously to reduce and prevent traffic crashes and make roadways safer. Secondary crashes are collisions that have taken place inside of incident scene or within the queue, influenced by the already occurred primary incident. Secondary crash occurrences are less frequent than primary incidents, however, the incident management without considering a potential secondary crash can cause a worst-case scenario to both emergency vehicles and travelers. Although statistical models have been developed in the past to estimate the probability of secondary crashes, they do not consider time-series changes of the probability. A Markov chain, a stochastic model is used in this study to model randomly occurring incidents in a sequence. It is different from previously developed semi-Markov models by considering incident duration as a first-order to estimate the secondary crash parameters in the second order. Based on author's previous models on incident duration prediction, this study develops a multivariate second-order Markov model to estimate the probability of a secondary crash based on various primary incidents. This analysis will determine and identify if the probability of a secondary crash is higher at a specific location or higher due to a specific type of primary incident. Findings from our analysis can aid in developing countermeasures such as allowing emergency operators to allocate more resources to clear primary incidents quicker, or better prepare for secondary crashes based on the predicted probability of additional incidents. [ABSTRACT FROM AUTHOR]