Ensemble Tree Approach to Estimating Work Zone Capacity

Abstract

Accurate estimation of work zone capacity is necessary for successful traffic control management at work zones. This study uses an ensemble tree approach to estimate work zone capacity more accurately than would a model that was based on a single decision tree. A bootstrap aggregation method is used to build an ensemble tree comprising a set of individual decision trees. In this method, a set of bootstrap samples is generated by sampling with replacement from a training sample. A set of individual trees is then constructed with a tree learning algorithm and combined through averaging of the output. Data on work zone capacity from 14 states and cities are used in a case study to build and evaluate the ensemble tree. The results of the statistical comparison demonstrate that the ensemble tree outperforms the existing models of work zone capacity in estimation accuracy. The ensemble tree also performs better than any single decision tree for stability. A comparison with the 2010 Highway Capacity Manual indicates that the ensemble tree can provide a more accurate estimate of work zone capacity. Unlike the manual's model, the ensemble tree avoids estimated errors caused by subjective judgments of users because it does not require manual setting of various adjustment factors. Because of its high estimation accuracy and stability, the ensemble tree is a good alternative for estimating work zone capacity, especially for inexperienced users.