Maritime Fuel Price Prediction of European Ports using Least Square Boosting and Facebook Prophet: Additional Insights from Explainable Artificial Intelligence

Abstract

Prediction of bunker fuel spot prices at a port and understanding the dependence on key determinants is an arduous and challenging activity. The present work strives to analyze the temporal spectrum of daily spot prices of Very Low Sulphur fuel Oil (VLSFO), a critical bunker fuel, in five European Ports, Amsterdam, Antwerp, Gothenburg, Hamburg, and Rotterdam. The lack of prior research in the allied domain has motivated to undertake the modeling of VLSFO spot prices through the lens of applied predictive analytics. The Least Square Boosting (LSBoost) and Facebook Prophet algorithms are used to draw forecasts in multivariate framework leveraging constructs related to the same fuel prices at different ports, different fuel prices at the same ports, economic indicator, etc. The dynamics have been explicitly examined during the Russia-Ukraine military conflict. Additionally, Explainable Artificial Intelligence (XAI) frameworks have been used to demystify the influence of the chosen explanatory variables at a granular scale. The overall findings espouse the effectiveness of the predictive framework in accurately estimating spot prices of VLSFO in any of the selected ports, and the same heavily depends on VLSFO prices at different ports.