A Deep Machine Learning Approach for Predicting Freeway Work Zone Delay Using Big DataAuthor/Presenter: Shabarek, Abdullah
The introduction of deep learning and big data analytics may significantly elevate the performance of traffic speed prediction. Work zones become one of the most critical factors causing congestion impact, which reduces the mobility as well as traffic safety.
A comprehensive literature review on existing work zone delay prediction models (i.e., parametric, simulation and non-parametric models) is conducted in this research. The research shows the limitations of each model. Moreover, most previous modeling approaches did not consider user delay for connected freeways when predicting traffic speed under work zone conditions. This research proposes Deep Artificial Neural Network (Deep ANN) and Convolution Neural Network (CNN) traffic speed prediction models, for upstream freeway segments, including those on connected freeways, under work zone conditions.
The developed models are able to identify the congestion on the connected links in addition to the upstream mainline segments. The models predict traffic speed with work zone conditions based on traffic volume approaching the work zone, speed during normal conditions, work zone capacity, distance from work zone, vertical road gradient, downstream traffic volume and type of freeway segment. Moreover, the previous efforts in non-parametric approaches did not consider a solution to the overfitting problem of Artificial Neural Network (ANN). The proposed Deep ANN and CNN models use a dropout regularization to mitigate the overfitting issues. When comparing the CNN model ii to the Deep ANN model and the results of the Work Zone Interactive Management APplication-Planning (WIMAP-P) model, the testing results show higher accuracy with the CNN model compared to the other two models. The CNN model has filters that extract useful inputs from previous layers and reduces the overfitting problems. Dropout regularization technique is used to prevent the co-adaptation of training data. The CNN model is calibrated by varying the number of neurons at each hidden layer, the number of hidden layers, the optimizer algorithm, the filter height and the filter stride. The results indicate that the CNN model outperforms Deep ANN and the model of WIMAP-P in predicting traffic speed under work zone conditions.
While traditional efforts were conducted previously on predicting traffic congestion on the upstream freeway segments, the developed CNN model helps transportation agencies in planning for work zones by including both connected freeways and the upstream segments when predicting traffic speed under work zone conditions. Therefore, transportation agencies can prepare more accurate congestion mitigation plans, and provide more accurate user delay plans.