Predicting the changes in international crude oil trade relationships via a gravity heuristic algorithm

Changes in international crude oil relationships significantly impact the stability of energy markets and the dynamics of global politics. Predicting these changes can help evaluate energy market risk and provide essential insights for countries to adjust their economic strategies, energy policies, and diplomatic approaches. Previous studies on the prediction of global crude oil trade relationships have relied primarily on undirected networks, focusing solely on establishing potential relationships and neglecting trade directions and cancellations. Therefore, this paper proposes a gravity variational graph autoencoder (G-VGAE) model to predict changes (establishment and cancellation) in trade relationships in the directed network of the international crude oil trade. The method considers node characteristics such as GDP, import/export volumes, in/out degrees, and political risk indices, as well as network structural features. We used 2017 and 2020 as benchmark years to evaluate the model's accuracy and used 2022 trade data for predicting relationship establishment and cancellation. The study found that incorporating node attributes improves model accuracy, with higher accuracy in predicting relationship cancellation compared to relationship establishment. Finally, based on the prediction results, this paper presents targeted policy recommendations to help countries formulate future crude oil trade strategies, promoting the sustainable development of the global energy market.