基于改进神经网络的无线网络流量预测  

Research on wireless network traffic prediction based on improved neural network

作　　者：卢敦陆[1] 张歆奕[2] 

机构地区：[1]广东省科技干部学院,广东广州510640 [2]五邑大学,广东江门529020

出　　处：《现代电子技术》2016年第10期30-33,共4页Modern Electronics Technique

基　　金：国家自然科学基金资助的课题(61120106002)

摘　　要：考虑到无线网络流量具有极强的分散性、随机性以及混沌等特性,使用传统的ARIMA预测模型和BP神经网络模型难以对其进行精确的预测等,该文使用粒子群优化算法对BP神经网络预测模型进行优化以解决BP神经网络容易陷入局部最小值以及训练收敛速率低等问题,引入遗传算法中的自适应变异因子来以一定概率初始化部分变量解决粒子群优化算法会出现陷入局部最优解以及早熟收敛等问题。最后使用经典的CRAWDAD数据库中的无线网络流量数据对该文预测方法性能进行测试,使用稳定小波变换方法将无线网络流量数据分解,得到由1个近似分量以及3个细节分量组成的数据流。测试结果表明,该预测算法在预测性能上要优于ARIMA预测模型和BP神经网络模型。Considering the characteristics of wireless network traffic,such as dispersion,randomness and chaos,the particle swarm optimization algorithm is used in this paper to optimize the prediction model of BP neural network to solve the problems that the BP neural network is easy to fall into local minimum and its training convergence rate is low because the traditional ARIMA prediction model and BP neural network model are difficult to predict accurately,in which the self-adaptive mutagenic factors in genetic algorithm are brought. The performance of the predictive method was tested by means of the wireless network traffic data in the classical CRAWDAD database. The stable wavelet transform method is used to decompose the wireless network traffic data to obtain the data flow composed of 1 approximate component and 3 detail components. The testing results show that predictive performance of the predictive method is better than those of the ARIMA predictive model and BP neural network model.

关 键 词：无线网络流量预测 粒子群优化算法 BP神经网络 ARIMA预测模型 

分 类 号：TN915-34[电子电信—通信与信息系统] TP393[电子电信—信息与通信工程]