Summary
Asia-Pacific Network Operations and Management Symposium
2022
Session Number:TS7
Session:
Number:TS7-01
VAE-TCN Hybrid Model for KPI Anomaly Detection
Bo Wu , Qian Xu, Zhenjie Yao, Yanhui Tu, Yixin Chen,
pp.-
Publication Date:2022/09/28
Online ISSN:2188-5079
DOI:10.34385/proc.70.TS7-01
PDF download
Summary:
Unsupervised anomaly detection for KPI(Key Performance Indicator) series has been an active research area due to its enormous potential for industry applications. Existing works have made extraordinary progress in KPI series representation, reconstruction and forecasting. However, long-term temporal patterns prohibit the model from learning reliable dependencies. To this end, we propose a novel approach based on VAE-TCN hybrid model. Our model utilizes both a VAE(Variational Auto-Encoder) module for forming robust local features over short windows and a TCN(Temporal convolutional network) module for estimating the long term correlation in the series on top of the features inferred from the VAE module. Extensive experiments on various public benchmarks demonstrate that our method has achieved the state-of-the-art performance.