Book Description

This dissertation investigates the potential of adopting spatial-temporal data and machine learning techniques to predict traffic speed for transportation networks. Traffic data, along with historical weather information from multi regions located in the state of Ohio, were analyzed. Different spatial-temporal cases are generated based on the preprocessed traffic data along with various weather conditions. The first part of the dissertation investigates vehicles' speed variation patterns for different peak periods and different days of the week under congested and non-congested conditions in order to measure and understand the variability patterns. Different spatial-temporal cases are generated based on the preprocessed traffic data along with various weather conditions. Results showed a noticeable difference between rural and urban interstates in terms of speed patterns under normal and event conditions. "The second aim of the dissertation is to investigate the characteristics of speed distribution patterns under free-flow and recurrent congestion by fitting different distribution models. Results showed that the Normal, Burr, and t-location distributions could provide superior fitting performance compared to its alternative models under free-flow conditions" (Hussein et al., 2021). Lastly, the dissertation investigates the potential of adopting spatial-temporal data using machine learning techniques to predict traffic speed. Based on the obtained results, it was indicated that the support vector machine with radial bases kernel outperformed other models. Support vector machine model captured the drivers' speed patterns with the best prediction accuracy among all machine learning algorithms. The findings of this dissertation assist transportation planners and transportation agencies in visualizing the impacts of recurring and non-recurring congestion on arterial and freeways. Knowledge of travel speed distribution is one of the essential aspects of evaluating the performance of the transportation system, which results in improving the reliability of traffic parameters forecasting. Accurate traffic speeds prediction enables a smooth and effective daily operation for logistics and people transport on the transportation network.