Modeling and Predicting Traffic Accidents at Signalized Intersections in the City of Norfolk, Va
Maheshwari, Sharad, D'Souza, Kelwyn A., Academy of Information and Management Sciences Journal
This paper is an extension of the previously completed study of accident-patterns in the City of Norfolk (Maheshwari & D'Souza, 2006). The multiple-regression model developed in the previous study was based on variables related to intersection geometry. In this study, additional intersection factors are accounted for, which include speed limit, road signage, vegetation and traffic light data. Despite the expanded data set, many other factors like signal type, signal policies, road closures, road conditions, and condition of road signs which could possibly impact the traffic accidents, were not available at the time of the study. The motivation behind this research is based on the literature that indicates that the intersection topography/design factors and traffic management rules might contribute to the traffic accidents.
The objectives of this research were to develop comprehensive statistical exploratory and predictive models for intersections accidents in the City of Norfolk, VA. The research analysis was conducted in three phases. First, a linear regression model was developed using the same techniques applied in the previous study. This was done to establish a baseline model for a comparison of results. At the second stage, an exploratory data analysis technique (two-step cluster method) was used in which the study sample of 58 intersections was divided into two separate groups of clusters according to the type of roads meeting at the intersection arterial, collector and/or local roads. The first cluster consisted of the intersections between a major arterial road and a collector or local road, whereas the second cluster was made up of intersections of a major arterial road with another arterial or a large collector road. Two separate linear regression models were developed for each cluster.
An independent sample of 15 intersections was used for validation of these regression models. All three models, showed about 15% to 21% variation between actual and predicted accident rate values. In each case, however, the deviation between actual and predicted accident values was statistically insignificant. The second cluster deviation was the least, suggesting that the regression model for the intersections between major arterial roads or large collector roads had a somewhat better predictive power than the model for intersections between major arterial roads and collector or local roads.
The main objective of this research was to study the signalized intersections in the City of Norfolk to delineate intersection geometry, road signage and other design factors which may be contributing significantly to traffic accidents. This research project is an extension of the previously completed study on the accident-patterns in the same city in which a multipleregression model was developed on a selected set of intersections for the City. The City of Norfolk is one of the largest and oldest cities in the Hampton Roads region; and is home to roughly quarter million people. It is one of the most congested cities in the region by the population density. Furthermore, in 2006 the Hampton Roads had the highest crash incidents in the state compared to other regions on the basis of millions of VMT (vehicle mile traveled) (Nichols, 2007). The City of Norfolk contributed roughly 17% of those crashes in the region with annual traffic accident count of approximately 5,400. These data suggest that the traffic safety study could be useful to the City and to the Hampton Roads region.
There is evidence in the literature suggesting that road design factors could impact the traffic safety. Several highway engineering factors like lane widths, shoulder widths, horizontal curvature, vertical curvature, super-elevation rate, median and auxiliary lane were estimated and designed based on some traffic safety considerations. Additional factors like road signage, vegetation, line of sight of a traffic signal, horizontal and vertical curvature, and number of driveways close to an intersection have also been reported to have an impact on traffic safety. …