摘要：本研究结合现有港口吞吐量预测理论和我国港口的实际情况，通过分析我国沿海港口货物构成及影响，建立了中国沿海港口聚类-VAR分货类吞吐量预测模型，提高了我国沿海港口吞吐量预测的精度和准确度，从而为我国沿海港口投资规划等提供更可靠的参考依据。本模型主要特点有：一是首次从港口吞吐量货物构成的角度来分析和预测我国沿海港口吞吐量，解决现有文献主要采用港口吞吐量或选取GDP等少量宏观经济指标时间序列变量进行预测而导致信息挖掘不充分的弊端；二是在考虑沿海港口主要货物吞吐量之间的内在协调关系的前提下，借助聚类分析思想，对我国港口分货类吞吐量指标进行处理，最大限度的保留了信息，保证模型的预测准确性；三是本模型方便灵活，可以推广到单个港口或港口群吞吐量预测上，甚至可以推广到时间序列指标较多情况下的更广泛的预测问题。

Abstract: Considering the Chinese port present conditions and port throughput prediction theory, cluster-Vector Auto-Regression(VAR) sub-category of goods throughput prediction model is set up by analyzing Chinese coastal port cargo composition. It improves the prediction precision and accuracy for Chinese coastal port throughput, which provides more reliable reference for investment planning, etc. The main characteristics of this model are: Firstly, the capacity of Chinese coastal ports is analyzed and forecasted from the port throughput composition. It solves the pitfall that existing literature doesn’t mine information deep enough due to choosing port throughput or a small number of macroeconomic indicators, such as GDP time series prediction variables for the forecast. Secondly, with considering the coordinative relationship between major coastal port cargo category throughputs, the coastal port cargo category throughputs index is treated by using cluster analysis method. The maximum information is retained to ensure the accuracy of prediction model. Thirdly, the model is convenient and flexible, which could be extended to the throughput prediction for a single port or port group, or even extended to a wider range of the prediction problem possessing more time series indicators.