基于局域相关向量机回归模型的小尺度网络流量的非线性预测

发布时间：2018-11-27 14:56

【摘要】：基于非线性时间序列局域预测法与相关向量机回归模型,本文提出了局域相关向量机预测方法,并应用于预测实际的小尺度网路流量序列.应用基于信息准则的局域预测法邻近点的选取方法来选取局域相关向量机回归模型的邻近点个数.对比分析了局域相关向量机预测法、前馈神经网络模型与局域线性预测法对网络流量序列的预测性能,其中前馈神经网络模型的参数采用粒子群优化算法来优化.实验结果表明:邻近点优化后的局域相关向量机回归模型能够有效地预测小尺度网络流量序列,归一化均方误差很小;局域相关向量机回归模型生成的时间序列具有与原网络流量时间序列相一致的概率分布;局域相关向量机回归模型的预测精度好于前馈神经网络模型的与局域线性预测法的.
[Abstract]:Based on the local prediction method of nonlinear time series and the regression model of correlation vector machine, a prediction method of local correlation vector machine is proposed in this paper, and it is applied to predict the actual small-scale network traffic series. The local prediction method based on information criterion is used to select the number of adjacent points in the local correlation vector machine regression model. The prediction performance of local correlation vector machine (LRVM), feedforward neural network (FNN) and local linear prediction (LLP) for network traffic sequence is compared and analyzed. Particle swarm optimization (PSO) is used to optimize the parameters of feedforward neural network (FNN) model. The experimental results show that the local correlation vector machine regression model can effectively predict the small scale network traffic series, and the normalized mean square error is very small. The time series generated by the local correlation vector machine regression model have the same probability distribution as the original network traffic time series, and the prediction accuracy of the local correlation vector machine regression model is better than that of the feedforward neural network model and the local linear prediction method.
【作者单位】：济南大学信息科学与工程学院;山东省网络环境智能计算技术重点实验室;
【基金】：国家自然科学基金(批准号:61201428,61070130,61173079) 山东省自然科学基金(批准号:ZR2010FQ020,ZR2011FZ003) 山东省优秀中青年科学家科研奖励基金(批准号:BS2009SW003) 中国博士后科学基金(批准号:20100470081)资助的课题~~
【分类号】：TP393.06

【参考文献】