基于关联度网络流量预测的加权局域线性模型

用来预测混沌时间序列的传统加权局域模型一般用空间距离来定义邻近点的权重,当重构相空间嵌入维数增大时预测效果不是很理想。考虑了关联度对预测中心动力学行为的影响,提出用关联度来定义权重的方法,建立了一个用来预测网络流量新型的加权局域线性模型。模拟试验结果表明,和传统加权模型相比,当嵌入维数较高的时候,该模型能在较大程度上提高网络流量的预测精度。

Adding-weight local-region linear model of network traffic forecast based on degree of incidence

When the embedded dimension of reconstructive phase space increase,applying the traditional adding-weight local-region model,which the weight of neighbor phase points is generally determined by space distance to forecast the ehaotie time series,is not so satisfied.In the paper,taking the incidence-degree impaet on the dynamieal behavior of foreeast eenter point into account,a novel adding-weight local-region linear model for forecasting network traffie is created,with the weight of neighbor phase points defined by incidence-degree between neighbor phase points with forecast center point.The result of simulation shows the presented model can greatly improve precision of network traffic forecasting when the embedded dimension is high,compared with the traditional method.