Enhancing Global Logistics Forecasting: Leveraging Machine Learning Approaches for Predicting Import and Export Flows

Logistics is one of the most important sectors in which the global economy depends. There is no a doubt that artificial intelligence is an important tool in developing and enhancing logistics services. In this paper, we investigate the use of machine learning (ML) models for global logistics forecasting on import and export flows as one of the most consequential problems in international trade. The field has many challenges such as the changing nature of global trade patterns and sparsity/inconsistency of data across ship traffic, port operations and customs, to name a few. Furthermore, geopolitics and changing seasons in economies further contribute to unforeseen shocks, making predictions less effective over time. In addressing these challenges, we tried different ML models, and ExtraTreesRegressor and RandomForestRegressor showed better results by getting a much higher accuracy and smaller prediction errors in comparison with gradient boosting models (like XGBoost). By offering accurate trade flow forecasts, these models greatly improve port operations and mitigate supply chain risks. The research also provides an in-depth examination of the variables impacting trade volumes, enabling valuable recommendations for policy makers and the industrial players.