基于时空特征交叉融合的网络流量预测
Network traffic prediction based on spatial-temporal features cross fusion
黄冬妹 1宁芊1
作者信息
- 1. 四川大学电子信息学院,成都 610065
- 折叠
摘要
精确的网络流量预测对网络资源合理分配、提高通信质量有着重要作用.然而网络流量存在着复杂的时空依赖性,呈现高度非线性、复杂性,这给流量预测带来了困难.经过对现有的网络流量预测文献进行研究,分析网络流量的时间性质和空间性质,提出时空特征交叉融合的网路流量预测模型STCFusion.并在三个公开的数据集ABILENE、GEANT和CERNET进行充分的实验,实验结果表明提出的STCFusion有明显效果.
Abstract
Accurate network traffic prediction plays an important role in the rational allocation of network resources and the im-provement of communication quality.However,network traffic has complex spatial-temporal dependencies,presenting a high degree of nonlinearity and complexity,which brings difficulties to traffic prediction.After studying existing literature on network traffic pre-diction,analyzing the temporal and spatial properties of network traffic,a network traffic prediction model STCFusion based on the cross fusion of spatial-temporal features is proposed.And sufficient experiments were conducted on three publicly available datas-ets,ABILENE,GEANT,and CERNET,and the experimental results showed that the proposed STCFusion had significant effects.
关键词
网络流量预测/自注意力机制/时空特征Key words
network traffic prediction/self-attention mechanism/spatial-temporal features引用本文复制引用
出版年
2024
NETL
NSTL
万方数据