【摘要】 在网络预测算法中传统的预测几乎都没有考虑流量的自相似性和高斯性,仅仅利用最大Lyapunov指数进行计算机网络流量的混沌性检验,对网络流量的预测也仅仅是以计算得到的最大Lyapunov指数为前提,算法精度受限。提出一种改进的Wolf一步预测算法,对网络流量通过自相似的FGN(FGN,fractional gaussian noise)过程处理,得到替代原网络流量的新的序列,新的替代流量序列具有自相似性,从而进行预测。仿真结果准确检验了网络流量的混沌性,预测结果表明,改进的预测算法在略有缩短最大预报时间下,精度却高很多,预测的误差小于3%的点比例比原传统算法提高了20%以上。更多还原

【Abstract】 According to the nonlinearity and chaotic property feature of the network traffic flow, the prediction of the network traffic flow was going to be done precisely. An improved one step Wolf prediction algorithm was proposed with the(FGN) Fractional Gaussian Noise processing. The surrogate data of the original traffic flow data was gotten, and the new surrogate data series had the property of self-similarity for the prediction. The simulation result test the chaotic property precisely, and the prediction result show that the improved prediction algorithm has much high precision with the price of shortening the maximum prediction time a little. The position ratio of prediction error which less than 3% is increased 20% more than traditional algorithm.更多还原