Time Series Anomaly Detection Based on Hreg-VAE-LSTM

In the Industry 4.0, smart factories often face data anomalies during the collection and transmission of industrial time series data. To overcome this limitation, we propose a time series anomaly detection model based on Hreg-VAE-LSTM. The model leverages a Variational Autoencoder (VAE) module to capture local features within short time windows and employs the Hreg regularization method to mitigate the issue of data imbalance. Subsequently, a Long Short-Term Memory (LSTM) network is used to model the long-term dependencies in the sequence based on the features extracted by the VAE. This design enables the proposed algorithm to effectively detect anomalies across multiple temporal scales. Extensive experiments conducted on five real-world industrial datasets proved our model demonstrate the effectiveness and superiority of our model.