一种基于混沌特性的网络流量改进预测算法

高速网络中网络流量具有自相似特征，这种自相似性特征和混沌现象的吸引子有着密切联系。基于相空间重构理论，用网络流量混沌时间序列重构与原网络动力系统等距同构的相空间，通过计算网络关联维数、Kolmogorov熵和最大Lyapunov指数，证实网络流量具有混沌特性。分别采用基于Wolf原始算法和改进算法的最大Lyapunov指数方法，对网络流量进行了预测，并计算了最大可预报时间。仿真结果表明，基于Wolf改进算法的预测方法精度和可靠性高，从而为有效预防网络拥塞奠定了基础。

An Improved Prediction Method of Network Traffic Flow Based on Chaos Characteristics

High-speed network traffic flow has a self-similarity characteristic which keeps in close contact with the attractor of chaos system. A new method based on the reconstruction theory of phase space was presented to analyze network flow, and reconstruct a phase space which is equidistant and isomorphic to network dynamic system by use of time sequence of network flow. The fractal dimension, Kolmogorov entropy and the largest Lyapunov exponents of the reconstructed phase-space were calculated from the one dimensional time sequence of network flow, thereby demonstrating the chaos phenomena lied in Internet traffic. A prediction of traffic flow in high-speed network was performed, the maximum predictable time was computed by applying the method of largest Lyapunov exponents based on the Wolf scheme and improved Wolf scheme. The simulation result shows that the prediction method based on the improved Wolf scheme has higher accuracy and reliability, and lays a foundation for preventing the network from congesting.