Dynamic Adjustment of Mobile Ocean Freight Rates Based on Big Data

Abstract

Amidst the intensifying competition in the shipping industry and the ongoing digitalization of global trade, ocean freight rates—characterized as multivariate time series—are influenced by a complex interplay of factors including port network structures, market supply and demand dynamics, and transportation costs. Traditional static pricing strategies have proven inadequate in adapting to the rapidly evolving market conditions. Accurate freight rate forecasting has emerged as a critical prerequisite for enabling dynamic adjustment strategies. However, conventional time series models often fail to capture the spatial correlations among multiple entities. Existing graph neural network (GNN)-based approaches typically rely on either predefined static or dynamic graphs, which lack the capacity to effectively model the interactions between inherent static structures and evolving temporal dependencies in the time series data. In this study, the prediction of mobile ocean freight rates was investigated. The intrinsic data characteristics were first analyzed to uncover the coupling mechanism between static structural features and dynamic temporal patterns in time series. A GNN model based on multivariate time series was then proposed to automatically extract dependencies from both static and dynamic graphs through a datadriven graph learning module. An information interaction mechanism was designed to achieve deep fusion of the two types of graph structures, thereby addressing the subjectivity associated with manually defined graphs and the limitations of single-graph modeling approaches.