摘要
随着宽带卫星互联网业务种类和需求的不断增加,提高卫星互联网的服务质量已引起人们的广泛关注。为了减少各种业务流量源汇聚对卫星通信系统的影响,自相似流量的预测变得十分重要。该文根据网络流量的自身特性,在传统神经网络预测模型的基础上,采用了一种更为切合其性质的长短期记忆网络预测模型对其进行流量预测。通过对ON/OFF物理模型产生流量进行训练,得到结果。通过图像、数据、误差指标多方面对比,进行预测结果评价。仿真结果表明:长短期记忆神经网络预测模型可以实现网络流量的预测,精度较高。
With the continuous increase of the variety and demand of broadband satellite Internet services,improving the service quality of the satellite Internet has attracted extensive attention.In order to reduce the impact of the convergence of traffic sources of various services on the satellite communication system,the prediction of selfsimilar traffic becomes very important.According to the characteristics of network traffic,based on the traditional neural network prediction model,this paper uses a long-and short-term memory network prediction model which is more suitable for its nature to predict its traffic,gets results by training the traffic generated by the ON/OFF physical model,and evaluates the prediction results by comparing images,data and error indicators.The simulation results show that the long-and short-term memory neural network prediction model can achieve the prediction of network traffic,with high accuracy.
作者
李佳男
李琳琳
LI Jianan;LI Linlin(Shenyang Ligong University,Shenyang,Liaoning Province,110159 China;Shenyang Open University,Shenyang,Liaoning Province,110003 China)
出处
《科技资讯》
2023年第13期11-14,共4页
Science & Technology Information
关键词
长短期记忆网络
自相似性
流量预测
ON/OFF模型
拥塞控制
Long-and short-term memory network
Self-similarity
Traffic prediction
ON/OFF model
Congestion control
