Study on crash prediction models considering the effects of lane change manoeuvres

The purpose of this study is to develop a crash prediction model taking into account the observed number of lane change manoeuvres as an explanatory variable. The negative binomial (NB) regression technique is used, and various measures based on the Akaike's Information Criterion (AIC) are employed for model assessment. The results show that models based on crashes occurred within the observed segments are all better than the ones based on crashes occurred on the observed segments with an inclusion tolerance of 100m. One of the two best models, with Akaike weight of 0.19 (rank the second, and the first one is 0.40), takes the number of lane change manoeuvres as one of the explanatory variable. This implies that the number of lane change manoeuvres can be a candidate variable in predicting freeway crashes, and taking lane change manoeuvres into consideration is a possible way to get better crash prediction models.