VesselAI CO2 Emissions Prediction using ML algorithms
Using a unique dataset, comprising historical maritime data for different vessel voyages, including vessel type, voyage information and environmental factors, we employ decision trees, a number of regression, and artificial neural network algorithms for predicting Vessel CO2 emissions. Developed in the context of H2020 VesselAI project.
Performance evaluation is conducted using established metrics such as R2 (%), Mean Absolute Error, Normalized Root Mean Squared Error and Symmetric Mean Biased Error. Our first prediction outcomes reveal that the Extra Trees Regressor and Multi-layer Perceptron regressor algorithms outperform the other methods in terms of prediction accuracy demonstrating the least amount of error. This assets highlights the potential of machine learning in predicting maritime CO emissions and provides insights for further research, such as exploring alternative algorithms and incorporating real-time data Integration.