Scholarlink

In this paper, we present an anomaly detection algorithm that combines Transformer with Auto Encoder (AE) for time series data analysis. We first employ a preprocessing technique based on Empirical Mode Decomposition (EMD) and Savitzky-Golay (SG) filtering to reconstruct the original Key Performance Indicator (KPI) data, which helps to eliminate noise and improve the quality of the data. Then, a Transformer-AE network model is established, which effectively captures temporal characteristics without the need for labeled data. Anomaly detection is achieved by assessing the deviation between the model’s predicted data and the original data, leveraging the principles of normal distribution. Experimental results show that in the realistic production cloud environment of Wanye, the accuracy of anomaly detection for this algorithm has been improved by 2%.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced Time Series Anomaly Detection in Cloud Platforms: Leveraging Transformer and Autoencoder Synergies

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Liye Pang,
  • Haiming Su,
  • Lianfa Zhang

摘要

In this paper, we present an anomaly detection algorithm that combines Transformer with Auto Encoder (AE) for time series data analysis. We first employ a preprocessing technique based on Empirical Mode Decomposition (EMD) and Savitzky-Golay (SG) filtering to reconstruct the original Key Performance Indicator (KPI) data, which helps to eliminate noise and improve the quality of the data. Then, a Transformer-AE network model is established, which effectively captures temporal characteristics without the need for labeled data. Anomaly detection is achieved by assessing the deviation between the model’s predicted data and the original data, leveraging the principles of normal distribution. Experimental results show that in the realistic production cloud environment of Wanye, the accuracy of anomaly detection for this algorithm has been improved by 2%.