Abstract: There are features of relativity and sudden in the same moment of different date. In order to forecast the network traffic, this paper puts forward a related short ARIMA model. After the model sets up, the parameters is educed by using improved method. The parameters of the model can update automatically according to the changeable data. Experimental results shows that the related short ARIMA model can better describe relativity and self similarity, and makes higher forecast precision compared with AR and ARIMA model.

Key words: user behavior, traffic forecast, ARIMA model, time series, network traffic, short relativity

摘要： 不同日期同一时刻的网络流量存在相关性和突发性。为准确预测网络流量，提出一种短相关ARIMA模型。对模型定阶后，运用改进的建模方法推导模型参数，使参数随样本数据的变化而更新。实验结果表明，与AR模型和ARIMA模型相比，该模型能更好地描述网络的相关性和自相似性，预测精度较高。

关键词: 用户行为, 流量预测, ARIMA模型, 时间序列, 网络流量, 短相关性

CLC Number: 

• TP393.02