Abstract:Transformer-based models for anomaly detection in multivariate time series can benefit from the self-attention mechanism due to its advantage in modeling long-term dependencies. However, Transformer-based anomaly detection models have problems such as a large amount of data being required for training, standard positional encoding is not suitable for multivariate time series data, and the interdependence between time series is not considered. To address these limitations, we propose a novel anomaly detection method, named EdgeConvFormer, which integrates Time2vec embedding, stacked dynamic graph CNN, and Transformer to extract global and local spatial-time information. This design of EdgeConvFormer empowers it with decomposition capacities for complex time series, progressive spatiotemporal correlation discovery between time series, and representation aggregation of multi-scale features. Experiments demonstrate that EdgeConvFormer can learn the spatial-temporal correlations from multivariate time series data and achieve better anomaly detection performance than the state-of-the-art approaches on many real-world datasets of different scales.

What problem does this paper attempt to address?

The paper aims to address the problem of anomaly detection in multivariate time series data. Specifically, the study focuses on the following issues: 1. **Problems with existing Transformer models**: - Require a large amount of training data. - Standard positional encoding is not suitable for multivariate time series data. - Do not consider the interdependencies between time series. 2. **Challenges faced**: - Lack of clear definition of anomaly patterns. - Noise in the input data. - Increased computational complexity due to the length of the time series. - Difficulty in capturing the interdependencies between time series, which are often nonlinear and non-stationary. - Time series data annotation is difficult and costly. To address these issues, the paper proposes a new method called EdgeConvFormer, which combines Dynamic Graph Convolutional Neural Network (GCN) with the Transformer model. The main contributions are as follows: 1. **Temporal Embedding**: Utilizes the Time2Vec method to capture the inherent periodicity and complex patterns in multivariate time series. 2. **Spatio-Temporal Topological Relationships**: Introduces a dynamic graph convolution module, EdgeConv, to infer the spatio-temporal hierarchical topological relationships between sensors and to compensate for the Transformer's shortcomings in processing local information. 3. **Multi-level Feature Extraction**: By integrating EdgeConv with Transformer modules in a multi-layer structure, the model enhances their capabilities in a hierarchical, multi-scale manner, forming the EdgeConvFormer model. This approach improves and aggregates global and local feature representations at different scales. Experimental results show that EdgeConvFormer can learn spatio-temporal correlations from multivariate time series data and achieves better anomaly detection performance than existing techniques on various real-world datasets.

EdgeConvFormer: Dynamic Graph CNN and Transformer based Anomaly Detection in Multivariate Time Series

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

The Study on Multivariate Time Series Anomaly Detection Via Convolutional Variational Autoencoder

Disentangled Dynamic Deviation Transformer Networks for Multivariate Time Series Anomaly Detection

Dynamic Deep Graph Convolution with Enhanced Transformer Networks for Time Series Anomaly Detection in IoT

Unsupervised Anomaly Detection in Multivariate Time Series Through Transformer-based Variational Autoencoder

Attention Based CNN-LSTM Network for Anomaly Pattern Classification of Multivariate Time Series

Hybrid Graph Transformer Networks for Multivariate Time Series Anomaly Detection

Graph Transformer Network Incorporating Sparse Representation for Multivariate Time Series Anomaly Detection

A two-stage adversarial Transformer based approach for multivariate industrial time series anomaly detection

TCF-Trans: Temporal Context Fusion Transformer for Anomaly Detection in Time Series

LGAT: A Novel Model for Multivariate Time Series Anomaly Detection with Imporved Anomaly Transformer and Learning Graph Structures

An enhanced abnormal information expression spatiotemporal model for anomaly detection in multivariate time-series

Multivariate Time-Series Anomaly Detection via Graph Attention Network

DDMT: Denoising Diffusion Mask Transformer Models for Multivariate Time Series Anomaly Detection

Anomaly Detection for Multivariate Time Series Based on Contrastive Learning and Autoformer

SiET: Spatial information enhanced transformer for multivariate time series anomaly detection

DTAAD: Dual Tcn-Attention Networks for Anomaly Detection in Multivariate Time Series Data

DGTAD: Decomposition GAN-based Transformer for Anomaly Detection in Multivariate Time Series Data

Local-Adaptive Transformer for Multivariate Time Series Anomaly Detection and Diagnosis.