Abstract
<title>Abstract</title> <p>Due to changes in time and geography, urban traffic systems are large-scale, heterogeneous data environments with several issues. In this paper, provide a real-time urban traffic flow prediction system that combines big data analysis and graph convolutional networks. Create a directed space-time graph of the metropolitan road network to simulate both the topology and variations in traffic at various periods. Build a framework for complicated time-based pattern identification and non-Euclidean spatial dependency using spatial graph convolution and temporal convolution modules using all of the sensor data from a big metropolis. Experiments on a large multi-source dataset demonstrate that the suggested approach has good accuracy and stability in every section of the city under various operating conditions. According to the aforementioned research, the quality of urban transport services will be preserved, and this operating mode will still be accessible in some situations or because of insufficient data. In summary, a system of intelligent urban mobility management based on data is now possible due to the multiple positive empirical outcomes gained in this study.</p>