基于加权改进的AR模型的负载预测研究

Load Prediction Based on an Improved AR Model with Weighting 

摘要:

负载预测在故障管理中有着十分重要的作用,通过对CPU负载以及内存使用率的预测可以对系统进行实时监控,预知未来时间段资源的可用性,发出异常告警；文中提出一种加权改进的自回归模型,通过对最小二乘法求出的参数进行加权处理,结合时间序列分析理论,建立一个负载预测模型,用于CPU负载和内存使用率的预测；实验证明,对AR模型的参数进行加权的方法优化了参数估计,预测误差减小了60%~80%。

关键词:负载预测;时间序列;自回归模型;最小二乘法

Abstract:

Load prediction is very important in fault management, forecasting CPU load and memory usage can monitor the system in time, and it can know if the resource is available in the future, then give the malfunction alarm. This article came up with an improved AR model for load prediction, it based on the parameters of least square with weighting and combined with time series theory to build load prediction model. The model can predict CPU load and memory usage. The results prove that the method of weighting has optimized parameters and the prediction error has reduced 60%~80%.

Key words:load prediction; time series; AR model; least square method 