Deep Graph Library (DGL) port of Graph Neural Network-Based Anomaly Detection in Multivariate Time Series(AAAI'21) Requirements
Original implementation in Pytorch-Geometric.
It detects anomalies in multivariate time seires and have four steps
- Sensor Embedding
- Learning Graph Structure
- Using Graph Attention Network to forecast
- Calculate Deviation Scores to identify anomalous events.
Dataset is expert-labeled telemetry anomaly datafrom Mars Science Laboratory MSL.