农业云视频平台虚拟机负荷预测半监督偏最小二乘法模型（英文）

发布时间：2018-04-20 22:00

本文选题：负荷 + 预测　；参考：《农业工程学报》2017年S1期

【摘要】：为了优化基础设施资源的效率,农业云视频平台虚拟机布局算法需要了解虚拟机当前和未来的资源工作效率,尽可能准确地预知下一步工作,如服务部署,虚拟机的部署、迁移或停止。然而,通常在预测中使用的样本非常小,可用于分析的数据有限。因此,该文研究设计了一个考虑时间因素,基于小数据集学习的滑动窗口模型。此外,鉴于现有的预测算法仍然有很大的改进误差率的空间,该文中采用基于滑动窗口与最小二乘法和半监督学习的数学方法相结合,提出了一种半监督偏最小二乘法(semi-supervised partial least squares,SS-PLS)的方法来计算上述预测。该文中,分析了在虚拟机使用SS-PLS负荷预测的可行性和优势。试验结果表明,基于滑动窗口模型结合SS-PLS,使得预测精度有了显着的改善,即均方根误差为1.777 86,平均绝对误差是1.331 2,平均绝对误差百分比为0.238 36,三者的增量分别5.47%、6.37%、6.12%。该研究可为云平台中虚拟机资源管理和优化提供一种参考方法。
[Abstract]:In order to optimize the efficiency of infrastructure resources, the virtual machine layout algorithm of agricultural cloud video platform needs to know the current and future resource efficiency of virtual machine, and predict the next step as accurately as possible, such as deployment of service, deployment of virtual machine. To move or stop. However, the samples usually used in forecasting are very small and the data available for analysis are limited. Therefore, a sliding window model based on small data set learning is designed in this paper. In addition, since the existing prediction algorithms still have much room for improving the error rate, a mathematical method based on sliding window, least square method and semi-supervised learning is used in this paper. In this paper, a semi-supervised partial least SS-PLS method is proposed to calculate the above prediction. In this paper, the feasibility and advantage of using SS-PLS load forecasting in virtual machine are analyzed. The experimental results show that the prediction accuracy is significantly improved based on the sliding window model combined with SS-PLS.The mean square error is 1.777 86 the mean absolute error is 1.331 2 the average absolute error percentage is 0.238 36 and the increment of the three is 5.476.37 and 6.12 respectively. This research can provide a reference method for virtual machine resource management and optimization in cloud platform.
【作者单位】：中国农业大学信息与电气工程学院;农业部农业信息获取技术重点实验室;
【基金】：National Spark Program Key Project(2015GA600002)
【分类号】：TP302;S126

【相似文献】