Abstract
Time series anomaly detection is widely used to monitor the equipment sates through the data collected in the form of time series. At present, the deep learning method based on generative adversarial networks (GAN) has emerged for time series anomaly detection. However, this method needs to find the best mapping from real-time space to the latent space at the anomaly detection stage, which brings new errors and takes a long time. In this paper, we propose a long short-term memory-based variational autoencoder generation adversarial networks (LSTM-based VAE-GAN) method for time series anomaly detection, which effectively solves the above problems. Our method jointly trains the encoder, the generator and the discriminator to take advantage of the mapping ability of the encoder and the discrimination ability of the discriminator simultaneously. The long short-term memory (LSTM) networks are used as the encoder, the generator and the discriminator. At the anomaly detection stage, anomalies are detected based on reconstruction difference and discrimination results. Experimental results show that the proposed method can quickly and accurately detect anomalies.
Highlights
In recent years, with the development of the Industrial Internet, industrial big data has become an important research topic
To make the variational autoencoder (VAE)-generative adversarial networks (GAN) learn the temporal dependence of time series, we combine the VAE-GAN with long short-term memory (LSTM) by using LSTM as the encoder, the generator and the discriminator of VAE-GAN
In the LSTM-based VAE-GAN, the LSTM networks for the encoder, the generator and the discriminator have the same size with depth 1 and 60 hidden units
Summary
With the development of the Industrial Internet, industrial big data has become an important research topic. In the industrial production process, the behavior of machines always changes based on usage and external factors that are difficult to capture [23] Under such circumstances, it is difficult to predict the time series even in a few time steps, resulting in the time series anomaly detection method based on the prediction model being no longer applicable. We propose a LSTM-based VAE-GAN for time series anomaly detection, which effectively solves the above problems. The generator and the discriminator are jointly trained at the training stage, it is not necessary to calculate the best mapping from real-time space to the latent space at the anomaly detection stage. A novel anomaly detection method based on VAE-GAN is proposed to detect anomalies in times series data from sensors. The anomaly score consists of the reconstruction difference of the VAE part and the discrimination results of the discriminator, which makes it more able to distinguish anomalies from normal data
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
