Abstract:Maritime transport is a pivotal logistics mode for the long-distance and bulk transportation of goods. However, the intricate planning involved in this mode is often hindered by uncertainties, including weather conditions, cargo diversity, and port dynamics, leading to increased costs. Consequently, accurate estimation of the total (stay) time of the vessel and any delays at the port are essential for efficient planning and scheduling of port operations. This study aims to develop predictive analytics to address the shortcomings in the previous works of port operations for a vessels Stay Time and Delay Time, offering a valuable contribution to the field of maritime logistics. The proposed solution is designed to assist decision making in port environments and predict service delays. This is demonstrated through a case study on Brazil's ports. Additionally, feature analysis is used to understand the key factors impacting maritime logistics, enhancing the overall understanding of the complexities involved in port operations. Furthermore, we perform Shapley Additive Explanations (SHAP) analysis to interpret the effects of the features on the outcomes and understand their impact on each sample, providing deeper insights into the factors influencing port operations.
What problem does this paper attempt to address?
### Problems the Paper Attempts to Solve
This paper aims to optimize port operations through predictive analysis, specifically focusing on the dwell time and delay time of vessels at ports. The paper points out that maritime transport is a crucial logistics method for long-distance and bulk cargo transportation, but its planning is often affected by uncertain factors such as weather conditions, cargo diversity, and port dynamics, leading to increased costs. Therefore, accurately estimating the total dwell time and any delay time of vessels at ports is essential for efficient planning and scheduling of port operations.
The main contributions of the paper include:
1. **Defining the Problem**: Clarifying the lack of research on predictive analysis of port operations in the current literature.
2. **Method Comparison**: Conducting a thorough comparison of various methods for predicting and classifying the total dwell time and delay time of vessels at ports.
3. **Key Factor Discovery**: Identifying key factors affecting port operations through feature importance analysis.
4. **SHAP Analysis**: Performing SHAP (Shapley Additive exPlanations) analysis to understand the impact of these key factors on the output results, providing deeper insights for each sample.
### Background and Motivation
In global maritime logistics, the activities of vessels within ports, although often overlooked, are a crucial part of their journey. Traditional research usually focuses on the navigation of vessels on the high seas, while the complex operations, cargo handling, and operational details within ports receive less attention. However, the time dynamics, efficiency challenges, and optimization opportunities within ports require more in-depth research. Optimizing port operations is not only vital for promoting trade and commerce but also for accommodating the increasing flow of cargo and passengers, supporting economic competitiveness and resilience in the context of globalization.
### Methods and Data
The paper selected ports in Brazil as a case study due to Brazil's rich maritime history and strategic position in global trade. The dataset covers multiple features related to maritime cargo transport and port operations from November 2017 to December 2018, divided into four major categories: cargo characteristics, geographic characteristics, operational characteristics, and dwell time characteristics.
### Models and Evaluation
The paper employed tree-based methods, statistical methods, and deep learning methods in supervised learning for prediction and classification. The main models used include Random Forest, Extreme Gradient Boosting (XGBoost), Neural Networks, and ResNet. Evaluation metrics include Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²).
### Results and Discussion
The experimental results show that tree-based methods outperform statistical methods and deep learning methods in this task. Random Forest performed the best in predicting total dwell time and delay time, with the lowest RMSE values. Feature importance analysis and SHAP analysis revealed key factors affecting port operations, such as berth, cargo weight, and navigation type. These analyses help understand the complexity of port operations and provide support for decision-making.
### Conclusion and Future Work
The paper optimized port operations through predictive analysis, providing valuable insights that help improve the decision-making efficiency of port management. Future research can further explore the impact of more external factors (such as weather information) on port operations and how to apply these methods to ports in other regions.