Weighted Local Linear Model of Network Traffic Forecast Based on Entropy Coefficient
YU Zheng-wei · Jisuanji gongcheng · 2007
Applying entropy coefficient instead of space distance to determine the neighbor point with their weights in phase space,a novel weighted local linear model is created,this model can overcome the defect of the traditional weighted local model in which the neighbor points with their weights are generally determined by distance so that when it is used to forecast the chaotic time series with the high embedded dimension are not so effective.The result of simulation shows the presented model can greatly improve precision of network traffic forecast when the embedded dimension of network traffic time series is high,compared with the traditional method.