In this paper, a predictive model for vessel fuel consumption is developed, utilizing data analysis specific to LNG and HFO fuel types. The methodology encompasses data preprocessing, separation based on operational modes, division into training and test datasets, K-fold validation, and hyperparameter tuning using LightGBM. The model's predictions show that LNG outperforms HFO slightly, with both models achieving high accuracy as indicated by low error metrics and R2 scores above 0.94, demonstrating reliable adherence to dataset trends.