具有重尾特性自相似网络通信量的预测

提出了一种基于alpha-平稳分布过程的网络通信量模型，该模型能够刻画自相似网络通信量的长程相关性和重尾特性，具有参数简约，物理意义明确的优点，可以准确描述Bellcore Lab的实际踪迹BC-Oct89Ext．根据此模型，提出了一种在累积量约束条件下Fisher信息（FI）最小化的基于alpha-平稳新息（innovation）的FARIMA（fractional auto regressive integrated moving average）预测方法，它可以预测同时具有长程相关性与短程相关性的自相似网络通信量．这种基于alpha-平稳新息的FARIMA预测方法能够在无限方差意义下使预测偏差最小化，对实际踪迹BC-Oct89Ext的预测实验证明了该预测方法的准确性和可靠性．

Prediction of self-similar network traffic with heavy tailness

A self-similar network traffic model is presented based on alpha-stable processes which can capture the long range dependence and the heavy tailness of the self-similar traffic.The model is fitted to bursty trace BC_-Oct89Ext collected at Bellcore Lab,and it is parsimonious in the number of parameters which have direct physical meaning.According to the new self-similar traffic model,a FARIMA(fractional auto regressive integrated moving average) predictor based on alpha-stable innovation was proposed by the minimization of the fisher information(FI) under cumulant constraints.FARIMA predictor can forecast the self-similar network traffic with long range dependence and short range dependence simultaneously.The new predictor is capable of minimizing the prediction dispersion under the meaning of the infinite invariance.The prediction experimental results for the actual trace BC_-Oct89Ext showed that the new FARIMA predictor is accurate and reliable.