Traffic Prediction for New York City Based on Graph Neural Network

The real-time speed of each roadway segment in a network is essential in many urban traffic environments, from transportation route planning in freight transportation to congestion avoidance in advanced traveler information systems. Therefore, accurate real-time average road speed prediction is of utmost importance. Vehicle speeds and travel times can vary considerably in urban environments due to traffic congestion caused by accidents or adverse weather conditions. At another level, one can also predict the future of each road segment by the speed variation of the surrounding road segments and the impact of the surrounding road segments on that road segment. This paper presents a project to predict the real-time average speed of selected road segments in New York City using graph neural networks. It focuses on predicting road segment speeds using Movement data from Uber vehicles and road network data collected over a considerable period as input. In this paper, different machine learning and data mining methods are applied, and computational results are reported on actual data to demonstrate the accuracy of the obtained predictions and the model's superiority.