Research on Container Freight Rate Prediction for Transpacific Routes Based on Multivariate LSTM: An Empirical Analysis Integrating Port Congestion and Capacity Indicators

Abstract

Accurate prediction of transpacific container freight rates is critical for shipping companies to optimize resource allocation, manage risks, and maintain stable operations in a volatile market environment. To address the limitation that traditional linear models struggle to capture the nonlinear and dynamic dependencies of freight rates, this study proposes a data-driven prediction framework based on a multivariate long short-term memory (LSTM) network for forecasting the China Export Container Freight Index on the US West Coast route. The model innovatively integrates multi-source heterogeneous data, including port congestion index, weighted container capacity, average ship speed, Brent crude oil price, and policy uncertainty index, and introduces four-week lagged features to characterize the delayed effect of market responses. Weekly data from January 2020 to December 2022 are used for model training and testing. Experimental results show that the proposed LSTM model achieves favorable prediction performance on the test set, with root mean square error (RMSE) and mean absolute error (MAE) as the evaluation metrics. Feature importance analysis reveals that the port congestion index and weighted container capacity are the key drivers of freight rate fluctuations. This study demonstrates that combining deep time-series models with fine-grained operational indicators can effectively improve short-term freight rate prediction accuracy, providing support for data-driven decision-making in the shipping and logistics industry.