Machine Learning Based Moored Ship Movement Prediction

Several port authorities are involved in the R+D+i projects for developing management decision-making tools. We recorded movements of 46 ships Outer Port Punta Langosteira (A Coruña, Spain) from 2015 until 2020. Using this data, we created neural networks and gradient boosting models that predict six degrees freedom a moored vessel ocean-meteorological data ship characteristics. The best achieve, surge, sway, heave, roll, pitch yaw movements, 0.99, 0.95, 0.98 R2 training have 0.10 m, 0.11 0.09 0.9°, 0.11° 0.15° RMSE testing, all below 10% corresponding movement range. these with forecast weather conditions sea state characteristics berthing location, can several days advance. These results good enough to reliably compare models’ predictions limiting motion criteria safe working (un) loading operations, helping us decide location operation when stop operations more precisely, thus minimizing economic impact cargo unable operate.