摘要
针对现有弹性云服务器(elastic cloud server,ECS)未来请求量预测模型准确度低、稳定性差等问题,提出一种基于指数平滑的Stacking集成预测模型。以多个二次指数平滑模型作为基础模型,将线性回归模型作为集成模型对多组指数平滑预测值进行最终拟合;预测过程中使用多组二次指数平滑模型对ECS的历史请求时序进行构造集成模型训练数据集并加入平滑系数的动态优化。与传统单一模型的对比实验结果表明,该模型在实际云服务器请求量预测过程中具有更好的准确性和稳定性。
Aiming at problems of the existing predication model of the elastic cloud server(ECS)like low accuracy and poor stability of future request volume,a Stacking ensemble prediction model based on the exponential smoothing was proposed.The multiple second exponential smoothing model was taken as the basic model,and the linear regression model was regarded as the integrated model,so as to conduct the final fitting of multiple exponential smoothing prediction values.The multiple second exponential smoothing model was used during the process of prediction to construct an ensemble model with training data set in accordance with the time sequences of the request volume of the elastic cloud server,and the smoothing coefficients gained dynamic optimization was introduced.According to the contrast experiment with the traditional single model,it shows that the proposed model possesses better accuracy and stability in the prediction process of the request volume of the actual cloud server.
作者
石瑾挺
孙洁香
邬惠峰
SHI Jin-ting;SUN Jie-xiang;WU Hui-feng(School of Computer Science and Technology,Hangzhou Dianzi University,Zhejiang 310018,China;Beijing Research Industry of Automation for Machinery Industry Limited Company,Beijing 100120,China)
出处
《计算机工程与设计》
北大核心
2020年第2期432-439,共8页
Computer Engineering and Design
关键词
弹性云服务器
预测模型
指数平滑
线性回归
集成模型
elastic cloud server
prediction model
exponential smoothing
linear regression
ensemble model
