Use transformer encoder for KPI anomaly detection

Tao Zhang; ShuQiu Li; Jie Pu; JunHua Zhang; XiaoYan Wang

doi:10.1117/12.2635854

5 May 2022 Use transformer encoder for KPI anomaly detection

Tao Zhang, ShuQiu Li, Jie Pu, JunHua Zhang, XiaoYan Wang

Proceedings Volume 12245, International Conference on Cryptography, Network Security, and Communication Technology (CNSCT 2022); 122450B (2022) https://doi.org/10.1117/12.2635854
Event: International Conference on Cryptography, Network Security, and Communication Technology (CNSCT 2022), 2022, Sanya, China

Abstract

KPI anomaly detection has always been a very important task in AIOps (Artificial Intelligence for IT Operations). Operations Engineers judge whether the service is stable by monitoring various KPIs. However, there are many challenges to build a KPI anomaly detection system, and the existing anomaly detection models are not strong. Given the effectiveness of the Transformer model in a series of fields such as NLP and CV, it has attracted great interest from researchers. Therefore, in this work, we build an end-to-end system named KPI_TransformerE for KPI anomaly detection. In particular, for the Class Imbalance problem, we use Focal Loss to solve it. Finally, we test our method on public datasets. The experimental results show that the method can achieve an F1 score of 0.963 in the test set which proves the power of our model.

Citation Download Citation

Tao Zhang, ShuQiu Li, Jie Pu, JunHua Zhang, and XiaoYan Wang "Use transformer encoder for KPI anomaly detection", Proc. SPIE 12245, International Conference on Cryptography, Network Security, and Communication Technology (CNSCT 2022), 122450B (5 May 2022); https://doi.org/10.1117/12.2635854

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available