【摘要】 指出港口吞吐量时间序列的研究较多采用机器学习和数据挖掘的方法,此类"黑箱"方法一般较难直观反映时间序列的规律特征。为了更直观地分析港口吞吐量时间序列特征,利用可视图理论构建了港口吞吐量时间序列复杂网络,并对这些网络的特性进行分析,如度分布、小世界效应、等级结构等。结果表明,可视图度值能够较为准确地确定预测值所在的区间,且新预测周期度值越大,预测区间更接近于实际观测值。更多还原

【Abstract】 In this paper, it is pointed out that the researches on the time series of port throughput usually adopt the method of machine learning and/or data mining. Such "black box" methods generally fall short in visually reflecting the regularity of the time series. To study it more intuitively, the complex network of the time series of port throughput is constructed based on the visual graph theory, whose characteristics such as degree distribution, small world effect and hierarchical structure, etc., are then analyzed. The results show that the viewability value can determine more accurately the interval where the predicted value is located, and the higher the value of the new prediction cyclicity is, the closer the prediction interval is to the actual observation value.更多还原