基于深度学习的网络流量快速预测方法  

Fast prediction method of network traffic based on deep learning

作　　者：许迎晨 XU Yingchen(Zhejiang ShineMo Technology Co.,Ltd.,Hangzhou 310013,China)

机构地区：[1]浙江讯盟科技有限公司,浙江杭州310013

出　　处：《无线互联科技》2024年第21期119-122,共4页Wireless Internet Science and Technology

摘　　要：针对现行方法在网络流量快速预测中应用存在预测误差较高的问题,文章提出基于深度学习的网络流量快速预测方法。该方法利用相空间对收集的网络流量时间序列数据进行重构,描述网络流量数据非线性结构;利用维度变换函数将网络流量数据转换为二维数据并通过提取数据自相似性特征,提取网络流量自相似性数据;利用深度学习对网络流量时间特征提取并通过特征融合预测未来时间点网络流量,实现基于深度学习的网络流量快速预测。实验证明,设计方法在网络流量快速预测场景中RMSE不超过0.01,MAE不超过0.1,可以实现对网络流量的精准快速预测。In view of the problem of high prediction error in the current method of fast network traffic prediction,a fast network traffic prediction method based on deep learning is proposed.Using phase space to collect network traffic time series data reconstruction,describe the network traffic data non-linear structure,using the dimensional transformation function of network traffic data into two-dimensional data,and extract the data from the similarity data,using deep learning of network traffic time feature extraction,and through the feature fusion,network traffic prediction future network traffic fast prediction based on deep learning.The experiment proves that the design method can achieve RMSE of 0.01 and MAE of 0.1 in the network traffic fast prediction scenario,which can achieve accurate and fast prediction of network traffic.

关 键 词：深度学习 网络流量 预测 相空间 维度变换函数 自相似性 

分 类 号：TP393.07[自动化与计算机技术—计算机应用技术]