Shipping & Ports
Port Congestion and Container Freight Rates: RBF Neural Network Prediction Model Achieves 96% Accuracy
A new study uses RBF neural networks to predict SCFI freight rates under port congestion, achieving a model goodness of fit of 96%, and revealing the transmission effects of freight rates across transpacific routes under extreme congestion, providing a new early warning tool for the shipping industry.
Container shipping is the main artery of global trade, but the increasing frequency of port congestion is intensifying freight rate volatility, posing challenges to supply chain cost control. A latest study published in *Frontiers in Marine Science* used a radial basis function (RBF) neural network to achieve high-precision predictions of the Shanghai Containerized Freight Index (SCFI) under port congestion scenarios, with the model's goodness of fit reaching 96%. The study also verified the transmission effect of freight rates under extreme congestion conditions, providing a new risk warning tool for the shipping industry.
Research Data and Methods
The study was jointly conducted by researchers from Kyung Hee University, Chung-Ang University, and Dankook University in South Korea. The team collected daily congestion data from four major ports—Shanghai, Busan, Los Angeles, and New York—from January 1, 2016 to January 1, 2023, and matched them with the corresponding weekly SCFI indices. The study selected the SCFI composite index and two key trans-Pacific routes—Shanghai-Los Angeles (SCFI1) and Shanghai-New York (SCFI2)—as prediction targets.
The RBF neural network has strong nonlinear fitting capabilities and is suitable for handling the complex dynamic relationship between port congestion and freight rates. The researchers used this model for regression prediction of the SCFI and used time-lag correlation models to analyze the linkage characteristics among different route indices.
Key Findings
The results show that the RBF neural network model achieves a goodness of fit (R²) of 96% for the SCFI composite index, and 94% and 93% for SCFI1 and SCFI2, respectively. This indicates a strong correlation between port congestion indicators and freight rate fluctuations, and that the deep learning model can effectively capture their nonlinear changes.
More importantly, the time-lag correlation models confirmed that there is a lagged correlation between SCFI1 and SCFI2 during abnormal extreme congestion periods. This means that port congestion not only affects local freight rates, but may also produce cross-route transmission effects, i.e., freight rate fluctuations on one route can spill over to another route within weeks. This transmission mechanism reflects the tight coupling of the global container shipping network.
Supply Chain Impact
Port congestion directly reduces effective capacity, pushes up freight rates, and increases logistics costs for cargo owners and freight forwarders. Shipping companies can respond by adjusting routes, schedules, or surcharges, and accurate freight rate forecasts facilitate these decisions. The model provided by the new study enables shipping companies to predict freight rate trends more promptly, optimize capacity allocation in advance, and thereby reduce market risk.
For cargo owners, freight rate forecasts can assist budget management and contract negotiation. In the context of the linkage between port congestion and freight rates, companies can use the forecast results to adjust inventory and transportation plans, reducing the impact of supply chain disruptions.
Regional Impact
The four ports covered by the study are all core gateways in their respective regions: Shanghai Port and Busan Port are important container hubs in Asia, while the Port of Los Angeles and the Port of New York serve the U.S. West Coast and East Coast markets, respectively. Congestion at these ports directly affects the stability and freight rate levels of trans-Pacific trade routes.研究揭示的传导效应表明，单个港口的拥堵可能通过航运网络扩散，对全球贸易走廊产生连锁影响。在供应链高度重视时效与低库存的当前环境下，跨太平洋航线的运价波动已成为国际物流风险管理的焦点。
行业观点与意义
研究者指出，尽管SCFI是国际航运市场的重要风向标，但针对拥堵情境下的预测研究仍相对不足。传统预测多集中于波罗的海干散货指数（BDI），而集装箱运价受合同结构、市场供需等因素影响，建模难度更高。RBF神经网络的应用为集装箱运价预测提供了新的量化方法。
该研究结论强调，港口拥堵的大数据与深度学习相结合，可以为集装箱运输组织提供决策基础，帮助规划航运策略并缓解市场风险。这一路径也展示了人工智能在物流领域应用的前景。
未来展望
随着物流科技的发展，此类预测模型有望进一步集成到数字货运平台和运价指数系统中，为市场提供动态预警。结合实时港口拥堵数据、物联网追踪信息以及航道通行状况，模型精度有望继续提升。在苏伊士运河、巴拿马运河等关键通道面临不确定性时，这类工具将增强全球供应链的韧性。
此外，传导效应模型可用于识别港口网络中的风险传播路径，帮助港口管理者和航运公司提前制定应对措施。对于新兴市场港口建设和贸易走廊优化，该方法也具有借鉴意义。
结论
港口拥堵已成为集装箱运价波动的重要驱动因素。基于RBF神经网络的预测模型以96%的拟合优度验证了深度学习在航运市场分析中的有效性，并揭示了跨航线运价传导机制。在全球供应链日益复杂的背景下，数据驱动的运价预测将成为航运业和物流管理者不可或缺的决策工具。
Local source note · logisticsnews
logisticsnews frames this note through Shipping & Ports / Port capacity / Carrier networks: Shipping & Ports / Port capacity / Carrier networks explains the local editorial angle. dates, names and status changes still need checking; Source links should be opened before the summary is reused.