Prediction of container port congestion status and its impact on ship’s time in port based on AIS data

ABSTRACT

The growth of container throughput and the impact of COVID-19 have made container ports increasingly congested, leading to uncertain cargo transit times and high freight rates. Shipping parties want to know in advance the congestion of each port in the next phase to adjust their plans, but there are few studies involving congestion prediction, and this study hopes to make some additions. Three indicators are extracted from Automatic Identification System (AIS) data to represent the port congestion status for the sake of generalizability. And those indicators are used to predict port congestion status and further to predict container ships’ time in port using the eXtreme Gradient Boosting (XGBoost) and SHapley Additive exPlanation (SHAP) algorithms. The established models show that those indicators can improve prediction accuracy of port time compared to not considering port status. The findings also show that port congestion status contributes significantly to determining port time and makes it fluctuate by up to nearly 50 hours. The prediction results could help shipping companies to change their transportation schedules early to bypass ports with long waiting times or to arrange ships to enter the port earlier for delivering cargos on schedule.

KEYWORDS:

• Container port
• port status
• port time prediction
• XGBoost
• SHAP

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Random forest is a classifier with multiple decision trees mainly for regression and classification. The result of RF: MAE = 84.8, $R^2=0.998$. The input variables are ship length, width, draft and Deadweight ton, and the output variable is design TEU.

Additional information

Funding

This work was supported by the Ministry of Industry and Information Technology [grant number MC-202009-Z03].