Graph Neural Networks-driven Traffic Forecasting for Connected Internet of Vehicles

The great progress of wireless communications breeds the prevalence of connected Internet of Vehicles (CIoV). Naturally, the internal connections among active vehicles also act as an indispensable factor in traffic forecasting. Although much related research work has been conducted during the past few years, they mainly designed and/or developed a single specific forecasting model for this purpose. Existing work may acquire ideal effects in some specific scenarios, yet lacking good robustness to dynamic change of situations. To deal with such a challenge, a graph neural networks-driven traffic forecasting model for CIoV is proposed in this work, deduced as Gra-TF for short. In this paper, we regard the dynamics of traffic data as the situation of temporal evolution. With the assistance of ensemble learning, three typical graph-level prediction methods are employed to construct an integrated and enhanced the forecasting model. Such a design is able to take advantage of several methods to minimize the uncertainty in CIoV. Finally, we select a real-world dataset to build the experimental situation for further assessment. Numerical results indicate that the proposed Gra-TF improves the prediction accuracy by 20% to 30% compared with several baseline methods.