A Combined Model for Network Traffic Forecasting Based on Maximum Entropy Principle
Lin Tian-feng · Microelectronics & Computer · 2006
s:On the ground of the fact that the network traffic is influenced by many uncertain factors,the maximum entropy principle in information theory is introduced to construct the combined model for forecasting,in order to improve forecasting precision through information entropy theory.In this model,the all individual forecasting model is accepted or rejected according to the historical forecasting error,and the result of their is regarded as constraints,and then the probability distribution of forecasting values is educed through the maximum entropy principle.Application results in a real network shows that the combined model improves the forecasting precision compare with individual model,and makes the forecasting results more reasonable and impersonal.