考虑模糊神经网络和小波变换相结合的实际流量预测算法  

The prediction method of actual traffic based on fuzzy neural network and wavelet transform

作　　者：陈国彬[1] 张广泉[2,3] 

机构地区：[1]重庆工商大学融智学院,重庆400033 [2]苏州大学计算机科学与技术学院,苏州215006 [3]中国科学院计算机科学国家重点实验室,北京100080

出　　处：《四川大学学报（自然科学版）》2014年第5期921-926,共6页Journal of Sichuan University(Natural Science Edition)

基　　金：重庆市教委科学技术研究项目(KJ133103);江苏省自然科学基金项目(BK2011152);中国科学院计算机科学国家重点实验室开放课题(CSYSKF0908)

摘　　要：针对无线传感器网络流量预测误差较大的问题,基于模糊神经网络和小波变换提出了一种新的预测算法(State Prediction algorithm based on Fuzzy Neural Network,SPFNN).首先该算法利用α-稳定分布对实际流量进行刻画,并给出了满足该分布的判断依据;其次,通过结合模糊神经网络和小波变换来提高实际流量的预测精度;最后,结合OPNET和MATLAB进行联合仿真,深入研究了影响该算法的关键因素,并对比其它算法性能,结果发现SPFNN具有较好的适应性.In order to overcome the lager prediction error of wireless sensor network, based on fuzzy neural network and wavelet transform, a novel prediction algorithm （State Prediction algorithm based on Fuzzy Neural Network,.SPFNN） is presented. At first, the algorithm uses α-stable distribution to depict the characteristic of actual traffic, and gives the distribution judgments. Then, the prediction accuracy of actual traffic is improved by neural network and wavelet transform. Finally, a simulation was conducted to study the key influence factor of algorithm with OPNET and MATLAB. The results show that, compared to other algorithm, SP-α has better suitability.

关 键 词：预测 误差 α-稳定分布 模糊神经网络 小波 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]