Abstract: Aiming at the problem of low prediction accuracy of actual service flow,a novel State Prediction algorithm based on Wavelet transform and LFSN (SPWL) is proposed by Linear Fractional Stable Noise (LFSN) and wavelet transform． In this algorithm,the characteristic of LFSN distribution is defined,and the fitting method of actual data which is satisfied LFSN progress is given with discrete Fourier transform． The prediction accuracy is improved by fusing the results of LFSN model． Simulation is conducted to study the key influence factor of algorithm with NS2 and Matlab,such as time delay,dropping rate,as well as utilization rate． Results show that,compared to FARIMA algorithm,SPWL algorithm has high accuracy prediction,the prediction error of SPWL is 12． 83% ．

Key words: service flow, prediction accuracy, distribution characteristic, Linear Fractional Stable Motion(LFSM), wavelet

摘要： 针对实际业务流预测精度偏低的问题,结合线性分形稳定运动(LFSM)模型和小波变换提出一种新的业 务流预测算法(SPWL)。定义线性分形稳定噪声(LFSN)分布特征,利用离散傅里叶变换产生满足LFSN 过程的数 列,并给出实际业务流数据拟合方法。通过小波变换降低实际业务流的突发特性,同时融合LFSM 模型的预测结 果提高实际业务流的预测精度。基于NS2 和Matlab 进行仿真实验,结果表明,与FARIMA 算法相比,SPWL 算法 预测精度较高,其预测误差仅为12． 83% 。

关键词: 业务流, 预测精度, 分布特征, 线性分形稳定运动, 小波

CLC Number: 

• TP393