Time Series Anomaly Detection Based on Language Model
Weixia Dang, Biyu Zhou, Weigang Zhang, Songlin Hu · 2020
Energy-saving studies on edge computing generally depend on the analysis of monitoring data of the energy-consuming equipment. However, due to the dynamic nature of the deployment environment, monitoring data is prone to failure or errors, which brings challenges to energy-saving performance. In this paper, we propose a novel method based on language model to detect anomalies in the monitoring data. Unlike recent proposals, this method using a small amount of labeled historical data and performs better. Technically, the architecture of this method comprise of three parts: the input representation, the original Bert and additional output layer. Simulation results demonstrate that this method only needs a small amount of label data to train the model to obtain better results than the state-of-the-art work.