Ensemble machine learning model prediction and metaheuristic optimisation of oil spills using organic absorbents : supporting sustainable maritime

The study discusses the urgent necessity of environmentally friendly remedies to tackle the increasing frequency of oil spills. Conventional remedial oil leak techniques, such as mechanical recovery and chemical dispersants, may cause environmental damage and suffer from low effectiveness. Organic absorbents are environment-friendly and low-cost substitutes; however, their acceptance is hampered by inadequate performance and optimisation studies. To close this gap, our work combines metaheuristic algorithms with ensemble machine learning and suggests a hybrid technique for the precise prediction and improvement of oil removal efficiency. Using Random Forest (RF) and XGBoost models, high R2 values (RF: 0.9517–0.9559; XGBoost: 0.9760), minimal errors, and strong generalisation were obtained by predictive modelling. Operating conditions were optimised using Grey Wolf Optimisation (GWO), showing an optimal percentage of oil removed (POR) of 93.59%. Combining metaheuristics with machine learning ensures accuracy and practical results, tackling the complexity of oil spill control. By concentrating on organic absorbents, the study fits with worldwide sustainability initiatives and provides a useful foundation for actual implementation.

---

Opracowanie rekordu ze środków MNiSW, umowa nr POPUL/SP/0154/2024/02 w ramach programu "Społeczna odpowiedzialność nauki II" - moduł: Popularyzacja nauki i promocja sportu (2025).