基于改进的量子粒子群优化小波神经网络的网络流量预测

发布时间：2018-07-29 11:22

【摘要】：为了改善小波神经网络(WNN)进行流量预测的性能及避免量子粒子群算法(QPSO)搜索后期的早熟收敛缺陷,提出了一种改进的QPSO。该算法定义粒子群聚拢度,改进收缩—扩张系数使其表示为聚拢度的函数并服从随机分布,以使粒子群具有动态自适应性,避免陷入局部最优,并通过搜索使用WNN待优化参数编码位置向量的粒子群的全局最优位置来实现目标参数的优化,使用本算法优化WNN参数,建立了基于改进的QPSO优化WNN的网络流量预测模型。使用真实网络流量通过两组对比实验对其预测精度进行验证,证明了该方法的可用性。实验结果表明,该方法的预测精度优于WNN和QPSO-WNN方法。
[Abstract]:In order to improve the performance of wavelet neural network (WNN) in traffic prediction and to avoid the premature convergence defects of quantum particle swarm optimization (QPSO) algorithm, an improved QPSO is proposed. The algorithm defines particle swarm cohesion, improves the shrinkage and expansion coefficient to express it as a function of cohesion and obeys random distribution, so that particle swarm has dynamic adaptability and avoids falling into local optimum. The optimization of target parameters is realized by searching the global optimal position of particle swarm optimization (PSO) which encodes position vector with WNN parameters. Using this algorithm to optimize WNN parameters, a network traffic prediction model based on improved QPSO optimization WNN is established. The accuracy of the proposed method is verified by two sets of comparative experiments, and the availability of the method is proved. Experimental results show that the prediction accuracy of this method is better than that of WNN and QPSO-WNN methods.
【作者单位】：国家数字交换系统工程技术研究中心;
【基金】：国家“973”计划资助项目(2012CB315901,2013CB329104) 国家自然科学基金资助项目(61372121) 国家“863”计划资助项目(2013AA013505)
【分类号】：TP183;TP393.06

【参考文献】