基于LSTM与传统神经网络的网络流量预测及应用  

Network Traffic Prediction and Application Based on LSTM and Traditional Neural Networks

作　　者：王海宁 袁祥枫 杨明川[1] WANG Haining;YUAN Xiangfeng;YANG Mingchuan(China Telecom Corporation Limited Beijng Research Institute,Beijing 102209,China)

机构地区：[1]中国电信股份有限公司北京研究院

出　　处：《移动通信》2019年第8期37-44,共8页Mobile Communications

摘　　要：提出了一个基于长短期记忆循环神经网络与传统神经网络的网络流量预测模型。通过与传统机器学习流量预测方法相对比,验证了该模型在网络流量预测中的适用性与更高的准确性。将上述流量预测模型应用在基于软件定义网络技术的智能化承载网切片系统中,以提升网络资源利用率,并提供了实验室验证结果。This paper proposes a traffi c prediction model based on long short term memory(LSTM)and traditional neural networks.Compared with traditional machine learning-based traffi c prediction methods,experiments show that the proposed model has the applicability and higher accuracy in the network traffi c prediction.Finally,the above traffi c prediction model is applied in the SDN-based intelligent transport network slicing systems to improve the utilization of network resources,and the laboratory verifi cation results are also provided.

关 键 词：流量预测 神经网络 长短期记忆 网络切片 

分 类 号：TN929.5[电子电信—通信与信息系统]