Neural Networks-Driven Ensemble Learning for Accurate ETA Prediction in Maritime Logistics

The Estimated Time of Arrival (ETA) of vessels is critical for maritime logistics efficiency, which reduces operational delays and improves supply chain reliability. Previous studies have primarily focused on single-model predictions, neglecting the potential advantages of integrating multiple models to fully leverage their respective strengths. This study aims to build an Artificial Intelligence (AI)-driven approach utilizing Artificial Neural Networks (ANNs) and ensemble learning to improve ETA prediction accuracy. This study leverages historical Automatic Identification System (AIS) data which includes GPS coordinates, speed of the vessel movement, and historical ship records. The dataset used Random Forest, data cleaning, and normalization means for preliminary data processing. Several machine learning models are applied for prediction to determine the model with better performance. To enhance accuracy, an ensemble learning approach integrates multiple ANN models with Random Forest regression for improving predictions. Experimental results show that ANNs significantly outperform traditional models. The ensemble learning model integrates multiple ANN predictions using Random Forest, that further enhances accuracy and robustness. Hyperparameter tuning shows that n_estimators = 2 0 provides the best trade-off between complexity and performance. This study underlines the superiority of ensemble AI modeling for ETA prediction in maritime logistics.