时间序列模型系统运行分析  

Time Series Model System Run Analysis

作　　者：王庆国 卢宝林 段修立 路春阳 董新忠 Wang Qing-guo;Lu Bao-in;Duan Xiu-i;Lu Chun-yang;Dong Xin-zhong(CHN Energy Information Control Interconnection Technology Co.,Ltd.,Beijing 100032,China)

机构地区：[1]国能信控互联技术有限公司,北京100032

出　　处：《科学与信息化》2023年第11期49-51,共3页Technology and Information

摘　　要：作为持续自主学习的第一步,也是关键的一步,数据增强学习主要用来解决神经网络训练时没有足够的数据来最大化深层神经网络的泛化能力的问题。文章提出了一种智能增强学习方法,将用于训练的数据按一种最优的数据增强策略进行扩维。增强学习一般用于图像数据扩维,针对时空数据的扩维方法目前并不多见。在时序预测领域,也常常面临着没有足够的数据来进行学习的窘境,因此本文将数据增强引入时空数据来提升时序预测的精度具有积极意义。As the first and crucial step of continuous independent learning,data augmentation learning is mainly used to solve the problem that there is not enough data in neural network training to maximize the generalization ability of deep neural networks.This paper proposes an intelligent augmentation learning method,which expands the data used for training according to an optimal data augmentation strategy.Augmentation learning is generally used for image data expansion,and there are currently few expansion methods for spatiotemporal data.In the field of time series forecasting,there is often a dilemma that there is not enough data for learning,so it is of positive significance that this paper introduces data augmentation into spatiotemporal data to improve the accuracy of time series forecasting.

关 键 词：日志信息 系统异常 检测 

分 类 号：G63[文化科学—教育学]