基于小波神经网络的实际业务流预测方法  

Theactual traffic prediction method based on wavelet neural network

作　　者：陈珂[1] 张立君[2] 彭志平[1] 柯文德[1] 

机构地区：[1]广东石油化工学院计算机与电子信息学院,茂名525000 [2]北京印刷学院信息与机电工程学院,北京102600

出　　处：《四川大学学报（自然科学版）》2013年第3期484-488,共5页Journal of Sichuan University(Natural Science Edition)

基　　金：国家自然科学基金(61272382);广东省科技计划项目(2012B010100037);广东省自然科学基金(10252500002000001;S2012010009963)

摘　　要：针对无线传感器网络传输过程中容易受到噪音干扰的问题,提出了一种新的业务流预测算法AWNNP(Ant colony-based Wavelet Neural Network Prediction).该算法首先利用小波变换对业务流进行分解,并将其小波系数和尺度系数作为样本数据.其次,结合蚁群算法和神经网络来训练样本数据,采用小波模型重构进行重构,以此获得业务流的预测数据.同时,通过仿真实验对比,并分析了小波神经网络预测算法和BP神经网络预测算法,实验结果表明,AWNNP算法性能较优,其误差为16.21%.In order to mitigate the interference problem by noise in wireless sensor network, a novel prediction algorithm AWNNP （Ant colony-based Wavelet Neural Network Prediction） is proposed. In this algorithm, actual traffic is decomposed with wavelet transform, which wavelet coefficients and scale coefficients are seen as sample data. Then, sample data are trained by ant colony and neural network, and they are reconstructed with wavelet model to get prediction data. A simulation was conducted to study the accuracy between AWNNP and Wavelet Neural Network Prediction, as well as BP Neural Network Prediction. The results show that AWNNP has better performance, and the residual is 16.21%.

关 键 词：无线传感器网络 预测 小波 神经网络 蚁群 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]