This research investigates the efficacy of integrating traditional gravity models with deep learning methodologies to predict trade flows of environmental goods. By merging machine learning with the gravity model, this study aims to enhance the accuracy of import and export predictions across various sectors. Analysing a comprehensive dataset spanning a 20-year period and employing advanced machine learning techniques, the paper identifies key variables (GDP, distance, common language, landlocked, etc.) influencing trade predictions and assesses their contributions within the models. The findings reveal that machine learning-enhanced models significantly improve predictive accuracy over the standalone use of traditional gravity models. This highlights the potential of combining advanced computational techniques with conventional economic models to better understand and forecast environmental goods trade dynamics. This study not only contributes to theoretical advancements in trade prediction but also offers practical insights for policymakers aiming to facilitate trade in environmental goods, ultimately supporting global sustainability efforts.