基于CVAE-LSTM的服务器KPI异常检测

打开文本图片集

中图分类号：TP183；TP368.5 文献标志码：A DOI：10.12305/j.issn.1001-506X.2025.03.34

Server KPI anomaly detection based on CVAE-LSTM

SHEN Xiarunl，*，LIRuonan²，ZHANG Haotian³ （1.Beijing Institute of Aerospace Information，Beijing lO0854，China；2.Patent Examination Cooperation（Beijing）CenterofThePatentOffice，Beijinglooo7O，China; 3. Sino-German College of Applied Sciences at Tongji University，Shanhai 20l804，China）

Abstract：The anomaly detection of keyperformance indicator（KPI） is the basis of allaspects of Internet intellgent operation and maintenance，and is of great significancefor fault alarm and server security.The depth generation model has beenable to solve the problem of poor depth feature representationability of machine learning model，butitis insuficientintermsof theprocessingoftimeinformationinKPIdataand thecaptureof long-term information.For this reason，aKPIanomaly detection model basedonthecombination of conditional variational autoencoder（CVAE）and long-short term memory（LSTM） is proposed.With the powerful representation ability of CVAE network，time information isadded to deep autoencoder，and the long-term memoryability of LSTM is used to improve the long-term anomaly learning and processing abilityof the proposed model.The trained CVAE network isused to further train LSTM. Through the comparison experiment with other deep learning models on three open datasets，the experimental results show that the performance of the model in this paper is better than thatof theLSTMalone and some deep learning models with better results in terms of F1 value.

Keywords：key performance indicator（KPI）anomaly detection；conditional variational autoencoder （CVAE）；long-short term memory（LSTM） network；KPI；deep learning

0 引言

联网的应用与服务已经深入到人们日常生活中的方方面面，巨大的网络流量也带来一系列网络安全威胁和风险，互联网公司的日常运维和服务器安全迎来了巨大的挑战。（剩余17334字）