Study On Explanatory Model Of Contributory Factors To Work Zone Crash Injury Severity

With the continued growth of China’s highway mileage in recent years, the construction and maintenance of highway has become a key issue in transportation infrastructure management. The existence of work zone will disturb the highway traffic significantly and increase the traffic conflicts. It will lead to the serious traffic accidents. By analyzing the contributory factor of work zone accident, we can better understand the regularity of work zone accident and casualty. In this way we can help the traffic safety administrators and work zone construction companies make a better plan for work zone construction activity, and also take some necessary measures to reduce work zone accident rate and casualty rate.Based on Michigan’s M-94/I-94/I-94BL/I-94BR work zone crash data of from 2004 to 2014, we use the Multinomial Logistic Regression, Classification and Regression Tree (CART), Random Forest and Association Rules respectively, to build the models of contributory factors and different crash severity. By comparing the results of four models, we can identify the risky contributory factors of work zone casualty crashes, and explain the relationship of those contributory factors. And we can explore the regulation of work zone casualty crashes comprehensively and reliably. Using data visualization technique can also help us better understand the regulation behind the contributory factors to the work zone casualty crashes.By summarizing the results of four models, we found the work zone accidents with truck/bus, motorcycle/bicycle or pedestrian involved are more serious. The rear-end crash and single motor crash will more possibly lead to casualty. The roads with speed limit greater than 40 mph, construction activity on road, and the undivided roadway without access control are riskier in work zone accident. The hazardous actions including careless/reckless driving and unable to stop in assured clear distance may lead to serious work zone accident. The unlighted darkness and wet pavement under the rain/snow weather are also risky in work zone casualty accidents.Moreover, we analyzed and ranked the performance of all four models. Prediction accuracy:Random Forests>CART>Logistic Regression; Interpretability:Association Rules=CART> Logistic Regression> Random Forests; Relationship among variables: Association Rules>CART>Random Forests=Logistic Regression; Data visualization: Association Rules>CART>Random Forests.