文献类型：期刊文献

中文题名：一种基于多标记学习的网络流量预测算法

英文题名：Network Traffic Prediction Algorithm Based on Multi Label Learning

作者：陈文庆[1]

机构：[1]中国劳动关系学院计算机应用技术教研室

年份：2016

卷号：32

期号：4

起止页码：139

中文期刊名：科技通报

外文期刊名：Bulletin of Science and Technology

收录：CSTPCD、、BDHX2014、BDHX

语种：中文

中文关键词：网络流量;预测;粒子群算法

外文关键词：network traffic;;prediction;;particle swarm optimization

中文摘要：通过对网络流量的准确预测提高对网络的调控和监测水平,避免网络拥堵,确保网络畅通。传统的网络流量预测算法采用粒子群算法,在处理大规模的流量数据时,容易导致信息发散和易陷入局部极值点,流量预测准确度不高。提出一种基于多标记学习混合差分粒子群进化的网络流量预测算法。构建多分簇的无线网络流量数据传输模型,对网络流量进行时间序列分析,采用粒子群优化算法分别比较网络信息流中的频率波动是否相同,对相同的进行合并,基于自回归移动平均算法,进行粒子群信息链特征优选准则设计,采用多标记学习混合差分粒子群进化算法,把网络流量数据嵌入到内核空间的超球体中,进行离线阶段的网络流量预测优化。仿真结果表明,该算法对网络流量预测的精度较高,误差减少,具有较好的应用价值。

外文摘要：Through the accurate prediction of network traffic, improve the level of regulation and control of the network, avoid network congestion and ensure the network flow. The traditional network traffic prediction algorithm uses particle swarm optimization algorithm, which is easy to lead to the information divergence and easy to fall into local extreme points when dealing with large-scale traffic data. A network traffic prediction algorithm based on multi label learning hybrid differential particle swarm optimization is proposed. The traffic data transmission model of multi cluster is constructed, and the time series of network traffic is analyzed, and the frequency fluctuation of the network information flow is the same as that of the network. The simulation results show that the algorithm has good application value for the prediction of network traffic with higher accuracy and less error.