基于加权滑动平均的磁盘使用率预测模型
张宗华1,张海全1, 李师航2,牛新征3

1)南京南瑞集团公司流程与信息管理中心,江苏南京 211106; 2)西南财经大学经济信息工程学院,四川成都 611130; 3)电子科技大学计算机科学与工程学院,四川成都 611731

计算技术; 加权滑动平均模型; 磁盘使用率; 自相关和偏自相关系数法; 拉依达准则; 权重转移; 多新息递推最小二乘法

Disk usage prediction based on an improved weighted moving average method
Zhang Zonghua1, Zhang Haiquan1, Li Shihang2, and Niu Xinzheng3

1)Process and Information Management Center, NARI Group Corporation, Nanjing 211106, Jiangsu Province, P.R.China2)College of Economics and Information Engineering, Southwestern University of Finance and Economics, Chengdu 611130,Sichuan Province, P.R.China3)College of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu 611731, Sichuan Province, P.R.China

computing technology; weighted moving average model; disk usage; autocorrelation and partial autocorrelation coefficient method; Pauta criterion; transferring weight; multiple innovation recursive least squares

DOI: 10.3724/SP.J.1249.2016.01072

备注

为能提前做好扩容准备,提出一种改进的加权滑动平均(weighted moving average, WMA)模型,用以预测未来短期内磁盘的使用率. 针对磁盘使用率序列变化较为平缓、要求滞后较小的特性,采用自相关和偏自相关系数法对模型定阶,处理数据后,在不影响精度的前提下计算最小滞后值,并使用结合了拉依达准则的权重转移法来均衡权重,用多新息递推最小二乘法对参数进行更精确的估计,以提高预测的准确性. 通过Matlab仿真实验可知,该算法预测误差小,滞后性弱,与原始WMA模型相比,具有更好的预测效果.

This paper proposes an improved weighted moving average(WMA)model to predict the usage of disks in the near future. Considering the characteristics of showing gentle change in disk usage and the requirements with small lag, we firstly utilize the autocorrelation and partial autocorrelation coefficient method to determine the order of the model. After processing the series, the minimum lag can be calculated on the premise without affecting the accuracy. Additionally, weights transferring combined with the Pauta criterion is used to balance the weight. At last, we estimate the parameters by using the multiple innovation recursive least squares to improve the result of prediction. According to the simulation result by Matlab, this algorithm is proved to have less result errors and a smaller lag. It provides a better prediction effect as compared with the original WMA model.

·