摘要
在基于机器学习的流量预测算法中,详细研究了基于回归模型的预测算法,将机器学习算法引入到网络流量预测中,提出了不同的弱回归算子用来描述网络流量中的非线性特性。针对网络流量中的自相似特性,提出两种不同的机制,即用主成分分析作为预处理和为每一维特征保留一组权重分布;同时,针对实验中发现的过匹配现象提出一种自适应的权重更新准则。
Based on machine learning flow prediction algorithm, we had a detailed study of the prediction algorithm based on the regression model to introduce machine learning algorithm into the network traffic forecasting, and then put forward different weak regression operators to de-scribe the flow of network non-linear characteristics. In view of network traffic self-similarity, we presented two different mechanisms. We first used the principal component analysis as pre- processing and for each one-dimension characteristics we retained a group of weight distribution. Meanwhile, we presented an adaptive weight updating guidelines for the found overmatching phe- nomenon in experiments.
出处
《淮海工学院学报（自然科学版）》
CAS
2012年第1期34-38,共5页
Journal of Huaihai Institute of Technology:Natural Sciences Edition
关键词
机器学习
流量预测
回归模型
算法
主成分分析
machine learning
traffic prediction
regression model
algorithm
principal compo-nent analysis
