Large Pretrained Foundation Model for Key Performance Indicator Multivariate Time Series Anomaly Detection
In the realm of Key Performance Indicator (KPI) anomaly detection, deep learning has emerged as a pivotal technology. Yet, the development of effective deep learning models is hindered by several challenges: scarce and complex labeled data, noise interference from data handling, the necessity to cap...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Open Journal of the Computer Society |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10811835/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841533387999805440 |
|---|---|
| author | Xu Wang Qisheng Xu Kele Xu Ting Yu Bo Ding Dawei Feng Yong Dou |
| author_facet | Xu Wang Qisheng Xu Kele Xu Ting Yu Bo Ding Dawei Feng Yong Dou |
| author_sort | Xu Wang |
| collection | DOAJ |
| description | In the realm of Key Performance Indicator (KPI) anomaly detection, deep learning has emerged as a pivotal technology. Yet, the development of effective deep learning models is hindered by several challenges: scarce and complex labeled data, noise interference from data handling, the necessity to capture temporal dependencies in time series KPI data, and the complexity of multivariate data analysis. Despite recent progress in large models that show potential for handling complex, multidimensional tasks, the lack of extensive, high-quality datasets presents a significant barrier for directly training these models in KPI anomaly detection. This scarcity limits the models' ability to learn and generalize effectively within this specific domain. To overcome this, we propose an innovative approach to adapt fully pretrained large models from other domains to KPI anomaly detection, thereby mitigating data constraints and enhancing detection precision. Our approach involves adapting large models to anomaly detection tasks using patch operations and fine-tuning techniques, which significantly enhances the model's temporal dependency capture capabilities. Furthermore, to address the multivariate challenge, we introduce a novel feature extraction method based on channel independence to optimize information processing across multidimensional features. Additionally, we leverage frequency domain information to design a feature enhancement method, further boosting the model's detection accuracy. By integrating these innovative techniques, we have developed a large-scale KPI anomaly detection model named ViTSD. Empirical evidence from experiments on five benchmark datasets and two additional datasets demonstrates ViTSD's superior performance, outperforming existing models across various evaluation metrics. |
| format | Article |
| id | doaj-art-14394ee7767946e4866363266ce52799 |
| institution | Kabale University |
| issn | 2644-1268 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Open Journal of the Computer Society |
| spelling | doaj-art-14394ee7767946e4866363266ce527992025-01-16T00:02:23ZengIEEEIEEE Open Journal of the Computer Society2644-12682025-01-01617718810.1109/OJCS.2024.352121710811835Large Pretrained Foundation Model for Key Performance Indicator Multivariate Time Series Anomaly DetectionXu Wang0https://orcid.org/0009-0007-0006-7781Qisheng Xu1https://orcid.org/0000-0003-4141-5950Kele Xu2https://orcid.org/0000-0001-5997-5169Ting Yu3https://orcid.org/0009-0004-4451-2275Bo Ding4https://orcid.org/0000-0002-1236-8318Dawei Feng5https://orcid.org/0000-0002-7587-8905Yong Dou6https://orcid.org/0000-0002-1256-8934College of Computer Science and Technology, National University of Defense Technology, Changsha, ChinaCollege of Computer Science and Technology, National University of Defense Technology, Changsha, ChinaCollege of Computer Science and Technology, National University of Defense Technology, Changsha, ChinaCollege of Computer Science and Technology, National University of Defense Technology, Changsha, ChinaCollege of Computer Science and Technology, National University of Defense Technology, Changsha, ChinaCollege of Computer Science and Technology, National University of Defense Technology, Changsha, ChinaCollege of Computer Science and Technology, National University of Defense Technology, Changsha, ChinaIn the realm of Key Performance Indicator (KPI) anomaly detection, deep learning has emerged as a pivotal technology. Yet, the development of effective deep learning models is hindered by several challenges: scarce and complex labeled data, noise interference from data handling, the necessity to capture temporal dependencies in time series KPI data, and the complexity of multivariate data analysis. Despite recent progress in large models that show potential for handling complex, multidimensional tasks, the lack of extensive, high-quality datasets presents a significant barrier for directly training these models in KPI anomaly detection. This scarcity limits the models' ability to learn and generalize effectively within this specific domain. To overcome this, we propose an innovative approach to adapt fully pretrained large models from other domains to KPI anomaly detection, thereby mitigating data constraints and enhancing detection precision. Our approach involves adapting large models to anomaly detection tasks using patch operations and fine-tuning techniques, which significantly enhances the model's temporal dependency capture capabilities. Furthermore, to address the multivariate challenge, we introduce a novel feature extraction method based on channel independence to optimize information processing across multidimensional features. Additionally, we leverage frequency domain information to design a feature enhancement method, further boosting the model's detection accuracy. By integrating these innovative techniques, we have developed a large-scale KPI anomaly detection model named ViTSD. Empirical evidence from experiments on five benchmark datasets and two additional datasets demonstrates ViTSD's superior performance, outperforming existing models across various evaluation metrics.https://ieeexplore.ieee.org/document/10811835/Time-frequency analysislarge foundation modelpatch operationKey Performance Indicator (KPI) anomaly detection |
| spellingShingle | Xu Wang Qisheng Xu Kele Xu Ting Yu Bo Ding Dawei Feng Yong Dou Large Pretrained Foundation Model for Key Performance Indicator Multivariate Time Series Anomaly Detection IEEE Open Journal of the Computer Society Time-frequency analysis large foundation model patch operation Key Performance Indicator (KPI) anomaly detection |
| title | Large Pretrained Foundation Model for Key Performance Indicator Multivariate Time Series Anomaly Detection |
| title_full | Large Pretrained Foundation Model for Key Performance Indicator Multivariate Time Series Anomaly Detection |
| title_fullStr | Large Pretrained Foundation Model for Key Performance Indicator Multivariate Time Series Anomaly Detection |
| title_full_unstemmed | Large Pretrained Foundation Model for Key Performance Indicator Multivariate Time Series Anomaly Detection |
| title_short | Large Pretrained Foundation Model for Key Performance Indicator Multivariate Time Series Anomaly Detection |
| title_sort | large pretrained foundation model for key performance indicator multivariate time series anomaly detection |
| topic | Time-frequency analysis large foundation model patch operation Key Performance Indicator (KPI) anomaly detection |
| url | https://ieeexplore.ieee.org/document/10811835/ |
| work_keys_str_mv | AT xuwang largepretrainedfoundationmodelforkeyperformanceindicatormultivariatetimeseriesanomalydetection AT qishengxu largepretrainedfoundationmodelforkeyperformanceindicatormultivariatetimeseriesanomalydetection AT kelexu largepretrainedfoundationmodelforkeyperformanceindicatormultivariatetimeseriesanomalydetection AT tingyu largepretrainedfoundationmodelforkeyperformanceindicatormultivariatetimeseriesanomalydetection AT boding largepretrainedfoundationmodelforkeyperformanceindicatormultivariatetimeseriesanomalydetection AT daweifeng largepretrainedfoundationmodelforkeyperformanceindicatormultivariatetimeseriesanomalydetection AT yongdou largepretrainedfoundationmodelforkeyperformanceindicatormultivariatetimeseriesanomalydetection |