As critical maritime gateways for the global economy,ports function as major hubs for international trade and significant sources of carbon emissions.Moving from the limitations of traditional studies that rely on short-term data from individual ports,this paper establishes a long-term,multi-port analysis framework.It systematically integrates 10 key indicators across multiple dimensions,including port operations and regional economic development,to establish a scientific basis for the long-term dynamic forecasting of port carbon emissions and to provide forward-looking decision-making support for low-carbon port development.A combination of Principal Component Analysis (PCA) and Grey Relational Analysis (GRA) is employed to scientifically screen these indicators,eliminating redundant information and precisely identifying the most representative core indicators.Ridge regression is then applied to further optimize the indicator system,thereby avoiding multicollinearity issues and ensuring its validity and reliability.Subsequently,a Support Vector Regression (SVR) model is introduced for carbon emission prediction.This model is enhanced by incorporating the Global Malmquist-Luenberger (GML) index to dynamically reflect the profound impact of changes in port production efficiency on carbon emissions.A comparative analysis yields the following key findings:container port carbon emissions are closely related with factors such as the number of berths,cargo throughput,container throughput,and operational costs;a prediction model constructed based on these core indicators can effectively track carbon emission trends;incorporating production efficiency characteristics significantly enhances the predictive accuracy of the SVR model,enabling a more precise understanding of carbon emission patterns in port operations.This study provides robust theoretical foundations and practical support for low-carbon port management.