Petrochemicals are important value-added products derived from oil. Computational methods may help stakeholders to identify the future markets and suppliers, which is the aim of this research. Previous studies on relation prediction in the fields of global energy have considered connection establishment only in unipartite networks. This research improves and extends the application of link prediction for petrochemicals by identifying weak trade connections and modeling the relations with bipartite graphs to cover country-product relations. For this purpose, positive and negative link prediction algorithms were implemented after import and export data extraction and preprocessing of the global petrochemical trade data for the period from the year 2017–2019. Then, the results were verified computationally and experimentally. The algorithm achieved an AUC greater than 90% and precision values of up to 0.76 for 63 product HS codes for different countries. The comparison of the results to real-world data confirmed at least a quarter of the forecasts for trade establishment and more than half for cancellation. Furthermore, recent practical results certified prominent predictions such as new trade cancellations for African countries and the important role of Belgium in import and export. Finally, various suggestions were made to improve the prediction accuracy.
