Port Congestion and Container Shipping Rates: RBF Neural Network Prediction Model Achieves 96% Accuracy

How Does Port Congestion Affect Container Freight Rates? RBF Neural Network Prediction Model Reveals a New Path

Introduction

The global container shipping market continues to be disrupted by port congestion. From Asia to North America, extended vessel waiting times and reduced fleet turnover efficiency are directly reflected in the sharp fluctuations of container freight rate indices. For shipping companies, cargo owners, and traders, predicting freight rate trends in advance has become key to risk management. However, the dynamic relationship between port congestion and freight rates has long lacked precise quantitative tools.

A study recently published in *Frontiers in Marine Science* attempted to answer this question using a deep learning model. The study used congestion data from four major ports—Shanghai, Busan, Los Angeles, and New York—from 2016 to 2023, combined with an RBF neural network, to forecast the Shanghai Containerized Freight Index (SCFI) and verified the cross-market transmission effect of freight rates under port congestion.

Key Progress

The study used the SCFI as the core proxy indicator for container freight rates, covering both the composite index and two major trans-Pacific routes: Shanghai–Los Angeles and Shanghai–New York. The researchers collected daily congestion data from the four ports from January 1, 2016, to January 1, 2023, and constructed a high-precision forecasting model.

The results show that the RBF neural network model achieved a coefficient of determination (R²) of 96% in predicting the SCFI composite index, 94% for the Shanghai–Los Angeles route index, and 93% for the Shanghai–New York route index. This means the model can explain the vast majority of freight rate fluctuations, demonstrating strong practical value.

More notably, the study confirmed the "spillover effect" of the SCFI through time-lag correlation analysis. Under major congestion conditions, lagged correlations exist among freight rate indices on different routes—that is, a freight rate change on one route may transmit to another route after several days or weeks. This finding confirms the "chain reaction" of port congestion and also explains why regional congestion often triggers global freight rate fluctuations.

Supply Chain Implications

For global logistics operators, this research offers multiple insights. First, a highly accurate freight rate forecasting model can help liner companies respond more precisely to congestion cycles in capacity deployment and slot pricing. Second, cargo owners and freight forwarders can use such tools to lock in freight rates in advance and avoid the risk of sudden price spikes.

The study also shows that the impact of port congestion is not limited to a single port or route. When severe congestion occurs at the Port of Shanghai or Los Angeles, freight rate fluctuations on trans-Pacific routes spread to other routes. This means supply chain managers need to assess congestion risk from a network perspective, rather than focusing only on the immediate status of the origin or destination port.Moreover, the predictive model offers a new approach to risk quantification in the shipping market. Traditionally, freight rate forecasting has relied mainly on historical data and macroeconomic indicators, whereas this study incorporates real-time congestion indicators into the model, significantly improving forecast performance. This suggests to the industry that port operational data can serve as an important input variable for freight rate forecasting.

Regional Impact

The four ports selected in the study are located in Asia and North America, covering one of the world's busiest trade corridors—the Trans-Pacific route. Shanghai is one of East Asia's largest container ports by throughput, Busan is an important transshipment hub in Northeast Asia, and Los Angeles and New York represent the import gateways on the U.S. West Coast and East Coast, respectively. Congestion at these four ports directly affects the timeliness and cost of the Sino-U.S. trade corridor.

The study's results show that there is a time-lagged correlation between freight rates on the Shanghai–Los Angeles route and the Shanghai–New York route, indicating a time difference in the transmission effect of congestion between the U.S. East and West Coast ports. This provides a reference for shippers choosing between U.S. West Coast and East Coast routes: if congestion at West Coast ports intensifies, freight rates may gradually transmit to East Coast routes, putting cost pressure on alternative routes that originally appeared more favorable.

For Asian export-oriented economies, this study also has policy implications. Port congestion is not merely an operational issue; it is a systemic factor affecting trade costs and international competitiveness. By forecasting freight rates more accurately, governments and port authorities can allocate resources in advance and mitigate the trade shocks caused by congestion.

Industry Views

"Container freight rate volatility is a long-standing challenge for the shipping industry, and port congestion often amplifies this volatility." The research team pointed out that most existing forecasting methods focus on the historical trends of freight rates themselves and rarely explicitly incorporate congestion factors into models. This study attempts to fill that gap and provide decision support for container shipping organizations.

The paper's authors stated that the RBF neural network is well suited to this scenario because it can handle nonlinear relationships and fit complex patterns through radial basis functions. With sufficient data, the model can quickly adapt to market changes.

However, the study also acknowledges several limitations. For example, the model has not yet incorporated external variables such as geopolitical events and fuel price fluctuations, which could be expanded upon in the future. In addition, the measurement of port congestion is relatively simplistic, and the model does not yet distinguish the structural causes of congestion (such as terminal equipment failures or labor shortages), but the preliminary results already show application potential.

Future Outlook

As global supply chains place greater demands on resilience, the value of freight rate forecasting tools will continue to rise. Future research directions may include: incorporating real-time data from more ports into the model, introducing the impact of sudden events such as Red Sea detours and Panama Canal capacity restrictions, and developing more granular route-level forecasts.

For logistics companies, embedding such models into digital freight management platforms could help them make more scientific freight rate decisions in an increasingly uncertain market. At the same time, port operators can use the forecast results to coordinate berth arrangements with shipping companies in advance, reducing the likelihood of congestion.

ConclusionThe dynamic connection between port congestion and container freight rates has moved from theoretical discussion to a quantitative modeling stage. This study uses real data to demonstrate the effectiveness of RBF neural networks in freight rate prediction and reveals the cross-market transmission patterns of freight rates under congested conditions. Although the model still has room for improvement, it undoubtedly provides a valuable analytical tool for the global container shipping market.

At a time when the global supply chain faces numerous uncertainties, being able to predict freight rates more accurately not only helps enterprises reduce operational risks but also contributes to maintaining the stable operation of international trade.

---

Source: https://www.frontiersin.org/journals/marine-science/articles/10.3389/fmars.2025.1545471/full