DA-LSTM-VAE: Dual-Stage Attention-Based LSTM-VAE for KPI Anomaly Detection
To ensure the normal operation of the system, the enterprise’s operations engineer will monitor the system through the KPI (key performance indicator). For example, web page visits, server memory utilization, etc. KPI anomaly detection is a core technology, which is of great significance for rapid f...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-11-01
|
| Series: | Entropy |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1099-4300/24/11/1613 |
| _version_ | 1797468307051249664 |
|---|---|
| author | Yun Zhao Xiuguo Zhang Zijing Shang Zhiying Cao |
| author_facet | Yun Zhao Xiuguo Zhang Zijing Shang Zhiying Cao |
| author_sort | Yun Zhao |
| collection | DOAJ |
| description | To ensure the normal operation of the system, the enterprise’s operations engineer will monitor the system through the KPI (key performance indicator). For example, web page visits, server memory utilization, etc. KPI anomaly detection is a core technology, which is of great significance for rapid fault detection and repair. This paper proposes a novel dual-stage attention-based LSTM-VAE (DA-LSTM-VAE) model for KPI anomaly detection. Firstly, in order to capture time correlation in KPI data, long–short-term memory (LSTM) units are used to replace traditional neurons in the variational autoencoder (VAE). Then, in order to improve the effect of KPI anomaly detection, an attention mechanism is introduced into the input stage of the encoder and decoder, respectively. During the input stage of the encoder, a time attention mechanism is adopted to assign different weights to different time points, which can adaptively select important input sequences to avoid the influence of noise in the data. During the input stage of the decoder, a feature attention mechanism is adopted to adaptively select important latent variable representations, which can capture the long-term dependence of time series better. In addition, this paper proposes an adaptive threshold method based on anomaly scores measured by reconstruction probability, which can minimize false positives and false negatives and avoid adjustment of the threshold manually. Experimental results in a public dataset show that the proposed method in this paper outperforms other baseline methods. |
| first_indexed | 2024-03-09T19:05:34Z |
| format | Article |
| id | doaj.art-83403375ef6e49e28f3b827172af63aa |
| institution | Directory Open Access Journal |
| issn | 1099-4300 |
| language | English |
| last_indexed | 2024-03-09T19:05:34Z |
| publishDate | 2022-11-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Entropy |
| spelling | doaj.art-83403375ef6e49e28f3b827172af63aa2023-11-24T04:37:05ZengMDPI AGEntropy1099-43002022-11-012411161310.3390/e24111613DA-LSTM-VAE: Dual-Stage Attention-Based LSTM-VAE for KPI Anomaly DetectionYun Zhao0Xiuguo Zhang1Zijing Shang2Zhiying Cao3School of Information Science and Technology, Dalian Maritime University, Dalian 116026, ChinaSchool of Information Science and Technology, Dalian Maritime University, Dalian 116026, ChinaSchool of Information Science and Technology, Dalian Maritime University, Dalian 116026, ChinaSchool of Information Science and Technology, Dalian Maritime University, Dalian 116026, ChinaTo ensure the normal operation of the system, the enterprise’s operations engineer will monitor the system through the KPI (key performance indicator). For example, web page visits, server memory utilization, etc. KPI anomaly detection is a core technology, which is of great significance for rapid fault detection and repair. This paper proposes a novel dual-stage attention-based LSTM-VAE (DA-LSTM-VAE) model for KPI anomaly detection. Firstly, in order to capture time correlation in KPI data, long–short-term memory (LSTM) units are used to replace traditional neurons in the variational autoencoder (VAE). Then, in order to improve the effect of KPI anomaly detection, an attention mechanism is introduced into the input stage of the encoder and decoder, respectively. During the input stage of the encoder, a time attention mechanism is adopted to assign different weights to different time points, which can adaptively select important input sequences to avoid the influence of noise in the data. During the input stage of the decoder, a feature attention mechanism is adopted to adaptively select important latent variable representations, which can capture the long-term dependence of time series better. In addition, this paper proposes an adaptive threshold method based on anomaly scores measured by reconstruction probability, which can minimize false positives and false negatives and avoid adjustment of the threshold manually. Experimental results in a public dataset show that the proposed method in this paper outperforms other baseline methods.https://www.mdpi.com/1099-4300/24/11/1613KPI anomaly detectionVAELSTMattention mechanismadaptive threshold |
| spellingShingle | Yun Zhao Xiuguo Zhang Zijing Shang Zhiying Cao DA-LSTM-VAE: Dual-Stage Attention-Based LSTM-VAE for KPI Anomaly Detection Entropy KPI anomaly detection VAE LSTM attention mechanism adaptive threshold |
| title | DA-LSTM-VAE: Dual-Stage Attention-Based LSTM-VAE for KPI Anomaly Detection |
| title_full | DA-LSTM-VAE: Dual-Stage Attention-Based LSTM-VAE for KPI Anomaly Detection |
| title_fullStr | DA-LSTM-VAE: Dual-Stage Attention-Based LSTM-VAE for KPI Anomaly Detection |
| title_full_unstemmed | DA-LSTM-VAE: Dual-Stage Attention-Based LSTM-VAE for KPI Anomaly Detection |
| title_short | DA-LSTM-VAE: Dual-Stage Attention-Based LSTM-VAE for KPI Anomaly Detection |
| title_sort | da lstm vae dual stage attention based lstm vae for kpi anomaly detection |
| topic | KPI anomaly detection VAE LSTM attention mechanism adaptive threshold |
| url | https://www.mdpi.com/1099-4300/24/11/1613 |
| work_keys_str_mv | AT yunzhao dalstmvaedualstageattentionbasedlstmvaeforkpianomalydetection AT xiuguozhang dalstmvaedualstageattentionbasedlstmvaeforkpianomalydetection AT zijingshang dalstmvaedualstageattentionbasedlstmvaeforkpianomalydetection AT zhiyingcao dalstmvaedualstageattentionbasedlstmvaeforkpianomalydetection |