Article Details

Abstract

In operations and maintenance scenarios, the data of interest is referred to as Key Performance Indicator (KPI). Faced with challenges in KPI anomaly detection, such as predominantly unsupervised data types, uncertain metric value ranges, and relatively few outliers, Digital Intelligence Department of China Mobile Communications Group Co., Ltd., leverages AIOps to accomplish the following work. Firstly, an algorithm for detecting Change Rate and Value upper-Bound (CRVB) is proposed based on AIOps. Then, ten KPIs from two critical scenarios interface services and process services are selected from the IT operations of a provincial branch of the telecom operator. This algorithm, along with four other industry-standard anomaly detection algorithms known for their effectiveness, is applied to the daily KPI anomaly detection tasks of the telecom operator, yielding results. Finally, by comparing anomaly detection evaluation metrics, it is determined the relative strengths and weaknesses of the five algorithms in terms of quality and efficiency, along with their respective suitable KPIs. The proposed CRVB algorithm demonstrates significant advantages over other algorithms in anomaly detection tasks involving certain proportion-based KPIs.

Keywords

intelligent operation; operator; anomaly detection; KPI; kernel density estimate

References

1. [1] Zhu Haiqi Research on Anomaly Detection of Operations Data Based on Deep Learning [D]. Harbin: Harbin Institute of Technology, 2019
2. [2] Anjing, Aiping, Xusen, etc A method for intelligent fault diagnosis of rotating machinery based on one-dimensional convolutional neural network [J]. Journal of Nanjing University (Natural Science Edition), 2019, 55 (1): 10
3. [3] HEGAZI A G，SHANAWANY E E E，EL-HOUSSINY A S. Prediction of Liver Diseases Based on Machine Learning Technique for Big Data [J//OL].BMC Veterinary Research，2018：127[2025-10-23].https://doi.org/10.1186/s12917-023-03689-y.
4. [4] PARK J W，TUMANOV A，J I A N G A，e t al. 3Sigma: Distribution-based Cluster Scheduling for Runtime Uncertainty [C]//EuroSys '18: Proceedings of the Thirteenth EuroSys Conference. Porto：Association for Computing Machinery，2018（2）：1-17.
5. [5] BENJAMIN M A，RIGBY R A，STASINOPOULOS D M. Generalized Autoregressive Moving Average Models [J].Publications of the American Statistical Association，2003，98（461）：214-223.
6. [6] MUHTAROV P，K UTIEV I，CANDE R L. Geomagnetically Correlated Autoregression Model for Shortterm Prediction of Ionospheric Parameters [J].Inverse Problems，2002，18（1）：49-65.
7. [7] SZMIT M，SZMIT A. Usage of Modified Holt-Winters Method in the Anomaly Detection of Network Traffic: Case Studies [J/OL]. Journal of Computer Networks and Communications，2012：192913[2025-11-02].https://doi.org/10.1155/2012/192913.
8. [8] GAO J K，SONG X M，WEN Q S, et al. RobustTAD: Robust Time Series Anomaly Detection via Decomposition and Convolutional Neural Networks [J/OL]. arXiv:2002.09545 [cs.LG].[2025-10-06].https://doi.org/10.48550/arXiv.2002.09545.
9. [9] LU W，GHORBANI A A. Network Anomaly Detection Based on Wavelet Analysis [J]. EURASIP Journal on Advances in Signal Processing，2008，2009：1-16.
10. [10] CARONI C. Outlier Detection by Robust Principal Components Analysis.
11. [J]. Communications in Statistics-Simulation and Computation，2000，29（1）：139-151.https://www.tandfonline. com/doi/abs/10.1080/03610910008813606.
12. [11] GE N Z，WENG X，YANG Q Y. Unsupervised Anomaly Detection Via Two-dimensional Singular Value Decomposition and Subspace Reconstruction for Multivariate Time Series [J]. Applied Intelligenc，2022：1-17.
13. [12] LATECKI L J，LAZAREVIC A，POKRAJAC D. Outlier Detection with Kernel Density Functions [C]//Machine Learning and Data Mining in Pattern Recognition.Berlin.Leipzig：Springer，2007：61-75.
14. [13] MIAO D，QIN X，WANG W. Anomalous Cell Detection with Kernel Density-based Local Outlier Factor [J].China communications，2015，12（9）：64-75.
15. [14] XU D，WANG Y，MENG Y，et al. An Improved Dataanomaly Detection Method Based on Isolation Forest [C]//2017 10th international symposium on computational intelligence and design(ISCID).Hangzhou：IEEE，2017，2：287-291.
16. [15] PRIMARTHA R，TAMA B A. Anomaly Detection Usingrandom Forest: A Performance Revisited [C]//2017 International Conference on Data and Software Engineering (ICoDSE).Quebec City：IEEE，2017：1-6.
17. [16] POL A A，BERGER V，GERMAIN C，et al. Anomalydetection with Conditional Variational Autoencoders [C]//2019 18th IEEE International Conference on Machine Learning and Applications (ICMLA).Boca Raton：IEEE，2019：1651-1657.
18. [17] DU M，LI F F，ZHENG G N，et al. Deeplog: Anomalydetection and Diagnosis from System Logs Through Deep Learning [C]//CCS'17: 2017 ACM SIGSAC Conference on Computer and Communications Security.Dallas：Association for Computing Machinery，2017：1285-1298.
19. [18] Wang Su, Lu Hua, Wang Shuo, etc Research progress on KPI anomaly detection in intelligent operation and maintenance [J]. Telecommunication Science, 2021, 37 (5): 42-51
20. [19] Xu Ke, Xu Mingwei, Wu Jianping A review of research on distributed denial of service attacks [J]. Small and Micro Computer Systems, 2004, 25 (3): 337-346
21. [20] Chen Yongbin, Cai Fangkai, Wang Fei Application of Data Warehousing and Data Mining Techniques in Telecommunications CRM [J]. Communication and Information Technology, 2009 (3): 80-82
22. [21] Li Zhong, Jin Xiaolong, Zhuang Chuanzhi, etc A review of graph oriented anomaly detection research [J]. Journal of Software, 2021, 32 (1): 167-193
23. [22] BOJER C S，MELDGAARD J P. Kaggle Forecasting Competitions: An Overlooked Learning Opportunity [J].International Journal of Forecasting，2021，37（2）：587-603.
24. [23] He Qing, Li Ning, Luo Wenjuan, etc Overview of Machine Learning Algorithms under Big Data [J]. Pattern Recognition and Artificial Intelligence, 2014, 27 (4): 327-336

Funding

Large model training and application development for IT operations tasks - Phase II (R25113U7); Deep Eye Intelligent Operation and Maintenance Platform Development - Phase II (R25113U8)

About the authors

Song Jiahao (1994-), male, Han nationality, from Zhaoqing, Guangdong, intermediate engineer, master's student, research direction: intelligent operation and maintenance; Chen Lihua (1984-), male, Han ethnicity, from Changsha, Hunan Province, intermediate engineer, master's student, research direction: intelligent operation and maintenance; Chen Le (1982-), male, Han ethnicity, from Chongqing, Sichuan, China, senior engineer, master's student, research direction: big data, intelligent operation and maintenance; Zuo Jinhu (1983-), male, Han ethnicity, from Hanchuan, Hubei Province, senior engineer, master's student, research direction: intelligent operation and maintenance; Zheng Wei (1984-), female, Han nationality, from Baoding, Hebei Province, intermediate engineer, master's student, research direction: intelligent operation and maintenance; Ye Zichun (1993-), female, Han nationality, from Yancheng, Jiangsu, intermediate engineer, master's student, research direction: intelligent operation and maintenance; Ding Zewei (1991-), male, Han nationality, from Shantou, Guangdong, is an intermediate engineer and a master's student. His research direction is intelligent operation and maintenance.