This study proposes a prediction model for fuel consumption in ships using XGBoost and SHapley Additive exPlanation (SHAP) to explain the predicted values. Previous studies have relied solely on operational data from ships to develop prediction models, neglecting the incorporation of external weather data. However, recently, a method has been applied to increase accuracy by utilizing both operational and external weather data. Nonetheless, the reliability of the prediction results and the variables used in the prediction model implementation remained unexplained. To address these issues, XGBoost and SHAP were used in this study to develop the prediction model.
This study provides an introduction to the research background, scope, relevant regulations, and previous studies, as well as the research methodology. It also explains the data cleaning method for ships and verifies the prediction model's results. Additionally, it covers XAI-related theories and the prediction model for fuel consumption in ships using XGBoost, as well as the SHAP-based method for explaining variable influence. Finally, it discusses the final results of this research and proposes future research directions.