Hot-spot analysis of motorcyclist crashes involving fixed objects using multinomial logit and data mining tools

Abstract

There has been a significant increase in the number of motorcyclists in the past years. According to crash statistics, motorcyclist crashes involving roadside fixed objects and safety systems have a higher probability of resulting in a fatality and serious injury. Therefore, as the number of motorcyclists on the road increases, there is an urgent need to consider motorcyclist safety as an important factor when designing roadside safety systems. In this paper, we are developing a hot-spot detection methodology to identify the design characteristics of locations associated with severe motorcyclist crashes for improving roadside safety systems. For this purpose, we are using the multinomial logistic regression and data mining tools such as random forests and decision trees. The results of both data mining and regression analysis show that roadside safety devices and fixed objects have a significant impact on RwD motorcyclist crash severity. Additionally, roadway characteristics (horizontal and vertical curvature, lane width, urban-rural classification) and operational factors (traffic volume and posted speed limit) are found to be associated with the RwD motorcyclist crash severity. Hence, to identify the high-risk locations (i.e., locations with potential for improvement), these roadway elements need to be accounted for.