基于奇异谱方法的网络行为分析

运用数字信号处理中的奇异谱分析方法，讨论了如何提取网络流量的周期和趋势特征，并结合最大熵谱方法对吞吐量趋势做出预测，将该方法运用于中国教育和科研网CERNET华东（北）地区网的主干网，分析了若干天的网络流量行为特征，分析结果与使用情况相吻合，对主干的流量进行了预测，并将预测结果与实际情况做了比较，定义了评价预测准确性的公式，将预测的准确性与国际上具有代表性的成果进行多方面的比较，结果说明该方法准确性较好。

Study of network behavior based on singular-spectrum analysis

A method borrowed from digital signal processing is applied to identify and retrieve low frequency variability and trend components from history metrics. Combined with maximum entropy method, predictions of network throughput can be made. This arithmetic was applied to the backbone throughput of China Education and Research Network Eastern China(North) Regional network, and the results were consistent with the practice. Predictions of the backbone throughput were made and the results were compared with the practice. A function which can evaluate the veracity of prediction results is defined, the veracity of this arithmetic is compared with the results of other international research communities. It is shown that this arithmetic has a better veracity.