计算机工程与应用 ›› 2020, Vol. 56 ›› Issue (5): 186-193.DOI: 10.3778/j.issn.1002-8331.1812-0278

• 模式识别与人工智能 • 上一篇    下一篇

高斯过程混合模型应用于网络流量预测研究

李松,周亚同,池越,何静飞,张世立   

  1. 河北工业大学 电子信息工程学院,天津 300401
  • 出版日期:2020-03-01 发布日期:2020-03-06

Application of Gaussian Process Mixture Model on  Network Traffic Prediction

LI Song, ZHOU Yatong, CHI Yue, HE Jingfei,  ZHANG Shili   

  1. School of Electronics and Information Engineering, Hebei University of Technology, Tianjin 300401, China
  • Online:2020-03-01 Published:2020-03-06

摘要:

精准的网络流量预测可以避免网络崩溃,保证网络的流畅度。将高斯过程混合(GPM)模型应用于网络流量的多模态预测。对两段不同地区的网络流量序列进行多模态分析,将之通过归一化和相空间重构后生成样本集并输入GPM模型。采用分类迭代学习算法,利用后验概率最大化和似然函数实现模型参数学习。将GPM模型与支持向量机(SVM)、核回归(KR)、最小最大概率机回归(MPMR)和高斯过程(GP)等模型比较。通过对比均方根误差[(RMSE)]和决定系数[(R2)]评价指标,GPM模型的预测准确度要优于其他四种模型。说明GPM模型能够很好应用于网络流量预测,可以为网络管理者分配网络资源提供参考。

关键词: 网络流量, 预测, 高斯过程混合模型, 多模态

Abstract:

Accurate network traffic prediction can avoid network crashes and ensure network fluency. The paper uses the Gaussian Process Mixture(GPM) model for multi-modal prediction of network traffic. Firstly, the multi-modal analysis of the network traffic sequences in two different regions is carried out, and then normalized and phase-space reconstructed to generate a sample set and input into the GPM model. Finally, the classification iterative learning algorithm is used to realize the model parameter learning by using the posterior probability maximization and likelihood function. The GPM model is compared with models such as Support Vector Machine(SVM), Kernel Regression(KR), Minimum and maximum Probability Machine Regression(MPMR), and Gaussian Process(GP). By comparing the Root Mean Square Error(RMSE) and the decision coefficient(R2) evaluation index, the prediction accuracy of the GPM model is better than the other four models. The GPM model can be well applied to network traffic prediction and can provide reference for network administrators to allocate network resources.

Key words: network traffic, prediction, Gaussian process mixture model, multimodal