A Hybrid Machine Learning Model For Ship Speed ThroughWater: Solve And Predict (Proposals Track)
Zakarya ELMIMOUNI (ENSAE); Ayoub Atanane (UQAR); Loubna Benabbou (UQAR)
Abstract
This research proposes a hybrid model for predicting ship speed through water, addressing challenges in estimating GHG emissions from shipping while contributing to climate change mitigation. Predicting ship speed through water is a key element in calculating GHG emissions. However, few models address this prediction in a way that integrates both physical principles and machine learning. Our approach combines physical modeling with data-driven techniques to predict real ship speed through water in two key steps: "Solve" and "Predict". In the first step “Solve”, a differential equation is resolved to estimate speed through calm water. "Predict" step uses a machine learning model that incorporates maritime and meteorological conditions and historical data to improve speed predictions in real-world conditions. This hybrid approach leverages both physics-based knowledge and machine learning models to provide a more comprehensive solution for accurately predicting ship speed through water.