基于图卷积神经网络的大规模软件定义网络流量预测模型
Large-scale software-defined network traffic prediction model based on graph convolutional neural network
张国明1
作者信息
- 1. 广东工商职业技术大学计算机学院,广东肇庆 526060
- 折叠
摘要
为了提高大规模软件定义网络流量预测的准确率,研究基于图卷积神经网络的大规模软件定义网络流量预测模型.构建包含图卷积神经网络(Graph Convolution Neural Network,GCN)层、门控递归单元(Gating Recursive Unit,GRU)层及自注意力机制层的流量预测模型.通过GCN层与GRU层分别重构与更新网络流量的空间与时间特征;将两种特征共同输入自注意力机制层,经整合与加权平均运算后,获得网络流量预测值输出,实现大规模软件定义网络流量预测.实验结果显示,该模型可精准预测大规模软件定义网络流量,降低所应用网络的通信丢包率与通信延时,实现高质量高时效的网络数据传输,保障大规模软件定义网络的智能流量通信.
Abstract
In order to improve the accuracy of large-scale software-defined network traffic prediction,a large-scale software-defined network traffic prediction model based on Graph Convolution Neural Network(GCN)is studied.Build a traffic prediction model including GCN layer,Gating Recursive Unit(GRU)layer and self-attention mechanism layer,reconstruct and update the spatial and temporal characteristics of network traffic through GCN layer and GRU layer respectively,input the two features into self-attention mechanism layer together,and obtain the network traffic prediction value output after integration and weighted average operation,to achieve large-scale software-defined network traffic prediction.The experimental results show that the model can accurately predict large-scale software-defined network traffic,reduce the communication packet loss rate and communication delay of the applied network,achieve high-quality and time-efficient network data transmission,and ensure intelligent traffic communication of large-scale software-defined network.
关键词
图卷积神经网络/软件定义网络/流量预测/门控递归/注意力机制/时间特征Key words
graph convolution neural network/software defined network/traffic prediction/gating recursion/attention mechanism/time characteristics引用本文复制引用
出版年
2024
NETL
NSTL
万方数据