Decision Tree–Based Model for Estimation of Work Zone Capacity

Abstract

The ability to estimate work zone capacity accurately is imperative because accurate estimates are a key input to estimates of queue length and traffic delay in work zones. This paper aims to develop a decision tree–based model that considers 16 influencing factors to estimate freeway work zone capacity accurately. The F-test splitting criterion and the postpruning approach are employed to grow and prune the decision tree. Freeway work zone capacity data collected from 14 states and cities are used to train, check, and evaluate the decision tree–based capacity estimation model. Statistical comparison results demonstrate that the decision tree–based model outperforms existing short-term and long-term freeway work zone capacity estimation models, especially when the input values of influencing factors are only partially available for the existing models. A comparison with the Highway Capacity Manual (HCM) also indicates that the decision tree–based model can provide a more accurate estimate of freeway work zone capacity. From the decision tree–based model, traffic engineers can easily estimate work zone capacity for a given freeway work zone by tracing a path down the tree to a terminal node. Because of its accuracy and ease of use, the proposed decision tree–based capacity model is a good alternative for traffic engineers to use in estimating freeway work zone capacity. It is expected that the decision tree–based capacity model could be applied to the HCM chapter on freeway facilities.