加权Internet访问直径的预测行为分析

研究Internet物理量短期及长期预测问题.首先定义了加权Internet网络,然后形式化定义表征加权Internet的物理参量-加权Internet访问直径,接下来根据海量数据变化特点,以径向基函数神经网络进行拟合,并针对实验结果对加权Internet访问直径进行预测分析.由于加权访问直径演化极为复杂,文中传统的一维数学模型在长期预测中已不适用.因此,在长期预测方面,首先计算得到加权Internet访问直径的关联维数,证实其演化过程为混沌现象,存在奇异吸引子.最后根据关联维数及奇异吸引子相近空间混沌轨道运动特性,提出了基于四维微分方程组的加权访问直径长期预测的形式模型.数据表明,短期及长期模型对Internet访问直径的短期及长期预测均可有效表达.

An Analysis on the Forecasting of Weighted-Internet Traveling Diameter

Physical properties of Internet are discussed in this paper.Based on giant samples,weighted Internet is firstly defined,and a physical property of weighted Internet-weighted traveling diameter is defined,then according to features of samples,a model is put foreword.The model is trained and simulated by RBF neural networks.Since the development of weighted traveling diameter is very complex,the model put foreword in this paper is far from suitable while being used for a long-term forecast.To solve the problem of long-term forecast,correlation dimension of weighted traveling diameter is firstly calculated,which proves that the development of weighted Internet traveling diameter is a phenomenon of chaos and a strange attractor exists.Finally,based on correlation dimension and properties of chaos path near the strange attractor,a model of four-dimensional differential equations group is put foreword which is comparatively suitable for a long-term forecast.