Abstract:In recent studies, Lots of work has been done to solve time series anomaly detection by applying Variational Auto-Encoders (VAEs). Time series anomaly detection is a very common but challenging task in many industries, which plays an important role in network monitoring, facility maintenance, information security, and so on. However, it is very difficult to detect anomalies in time series with high accuracy, due to noisy data collected from real world, and complicated abnormal patterns. From recent studies, we are inspired by Nouveau VAE (NVAE) and propose our anomaly detection model: Time series to Image VAE (T2IVAE), an unsupervised model based on NVAE for univariate series, transforming 1D time series to 2D image as input, and adopting the reconstruction error to detect anomalies. Besides, we also apply the Generative Adversarial Networks based techniques to T2IVAE training strategy, aiming to reduce the overfitting. We evaluate our model performance on three datasets, and compare it with other several popular models using F1 score. T2IVAE achieves 0.639 on Numenta Anomaly Benchmark, 0.651 on public dataset from NASA, and 0.504 on our dataset collected from real-world scenario, outperforms other comparison models.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is unsupervised anomaly detection in time - series data. Specifically, time - series anomaly detection is a very common but extremely challenging task in many industries, and it plays an important role in areas such as network monitoring, facility maintenance, and information security. However, since the data collected in the real world usually contains noise and the anomaly patterns are complex and diverse, it is very difficult to detect anomalies in time - series with high precision. Existing methods, such as those based on statistical models and traditional machine - learning methods, have limitations when dealing with complex time - series anomaly patterns. For example, the S - H - ESD method is only applicable to periodic data, and One - Class SVM requires a large amount of feature engineering to achieve good performance. To solve these problems, the author proposes a new model based on Nouveau VAE (NVAE) - Time - Series - to - Image VAE (T2IVAE), which is an unsupervised anomaly - detection model for univariate time - series. The main innovations of T2IVAE are as follows: 1. **Time - series conversion to image**: Convert one - dimensional time - series into two - dimensional images as input, which can capture more time features and correlations. 2. **Adopting NVAE structure**: Utilize the deep hierarchical structure of NVAE to improve the reconstruction effect of time - series. 3. **Introducing GAN technology**: Apply Generative Adversarial Network (GAN) technology during the training process to reduce over - fitting. Through these improvements, T2IVAE can detect anomalies in time - series more effectively, especially when dealing with complex and noisy data. In the paper, the author evaluates the performance of T2IVAE on three datasets and compares it with several existing methods. The results show that T2IVAE achieves good results on multiple indicators.

NVAE-GAN Based Approach for Unsupervised Time Series Anomaly Detection

Time Series Anomaly Detection with Variational Autoencoders

StackVAE-G: an Efficient and Interpretable Model for Time Series Anomaly Detection

VELC: A New Variational AutoEncoder Based Model for Time Series Anomaly Detection

LARA: A Light and Anti-overfitting Retraining Approach for Unsupervised Time Series Anomaly Detection

Anomaly Detection of Time Series With Smoothness-Inducing Sequential Variational Auto-Encoder

CANet: Convolutional Attention-based Network for Multivariate Time Series Anomaly Detection

Self-adversarial variational autoencoder with spectral residual for time series anomaly detection

Multivariate time series anomaly detection with variational autoencoder and spatial–temporal graph network

Weakly Augmented Variational Autoencoder in Time Series Anomaly Detection

Disentangled Anomaly Detection for Multivariate Time Series

Revisiting VAE for Unsupervised Time Series Anomaly Detection: A Frequency Perspective

Variational Autoencoder for Anomaly Detection: A Comparative Study

VESC: a new variational autoencoder based model for anomaly detection

Anomaly Detection for Time Series Using VAE-LSTM Hybrid Model

Reconstruct Anomaly to Normal: Adversarial Learned and Latent Vector-constrained Autoencoder for Time-series Anomaly Detection

Unsupervised Anomaly Detection Using Variational Auto-Encoder based Feature Extraction

A Comparison of Supervised and Unsupervised Deep Learning Methods for Anomaly Detection in Images

Self-Supervised Variational Graph Autoencoder for System-Level Anomaly Detection

Memory-augmented Adversarial Autoencoders for Multivariate Time-series Anomaly Detection with Deep Reconstruction and Prediction

Anomaly detection in sensor data via encoding time series into images