Anomaly detection on multivariate time-series

List of papers & datasets for anomaly detection on multivariate time-series data.

Contents

• 1. Papers
• 2. Books
• 3. Datasets
• 4. Evaluate
  • 4.1 Metrics
  • 4.2 Threshold
    • 4.2.1 Label-Based Threshold Search
    • 4.2.2 Thresholds Search without labels
• 5. Point Adjust & Point Adjust %K & Original

1. Papers

<div style="text-align: center"> <img src="https://github.com/qiumiao30/time-series-anomaly-detection/blob/main/image/anomaly%20detection.png"/> </div>

| Name | Code | Key word | Published | |------|------|----------|---------| | A Deep Neural Network for Unsupervised Anomaly Detection and Diagnosis in Multivariate Time Series Data | MSCRED|CH2 | AAAI'19 | | Graph Neural Network-Based Anomaly Detection in Multivariate Time Series | GDN| CH2| AAAI'21 |
| Multivariate Time-series Anomaly Detection via Graph Attention Network | MTAD_GAT| CH2 | ICDM'20 | | USAD : UnSupervised Anomaly Detection on Multivariate Time Series | USAD| adversarial | KDD'20 | | MAD-GAN: Multivariate Anomaly Detection for Time Series Data with Generative Adversarial Networks | MAD_GAN | | ICANN'19 | | Robust anomaly detection for multivariate time series through stochastic recurrent neural network | OmniAnomaly | | KDD'19 | | Deep Autoencoding Gaussian Mixture Model for Unsupervised Anomaly Detection | DAGMM | | ICLR'18 | | TranAD: Deep Transformer Networks for Anomaly Detection in Multivariate Time Series Data | TranAD | | VLDB'22 | | Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy | Anomaly Transformer| | ICLR'22 | | Timeseries Anomaly Detection using Temporal Hierarchical One-Class Network Lifeng | THOC(None)| | NeurIPS'20 | | Unsupervised Deep Anomaly Detection for Multi-Sensor Time-Series Signals | CAE-M(None)| | TKDE'21 | | Learning Graph Structures with Transformer for Multivariate Time Series Anomaly Detection in IoT | GTA| | IoTJ'21 | | Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding | InterFusion| | KDD'21 | | Detecting spacecraft anomalies using lstms and nonparametric dynamic thresholding | LSTM-NDT| | KDD'18 | | Time Series Anomaly Detection for Cyber-physical Systems via Neural System Identification and Bayesian Filtering | NSIBF| | KDD'21 | | | | | | | | | | | | | | | | | | | | | | | | | |

2. Books

• Outlier Analysis
• Introduction to Time Series and Forecasting, Second Edition
• Anomaly Detection Principles and Algorithms
• Outlier Ensembles An Introduction

3. Datasets

• SWaT & WaDI: SWaT Dataset Download, SWaT Dataset Introduce, WaDI Dataset Introduce | data_preprocess
• MSL & SMAP: Dataset Download and Introduction | data_preprocess
• SMD: Dataset Download and Introduction | data_preprocess
• ASD: Dataset Download and Introduction
• PSM: Dataset Download and Introduction
• KDDCup99: Dataset Download and Introduction
• MSDS: Dataset Download and Introduction
• MIT-BIH: Dataset Download and Introduction
• KDDCup21: Dataset Download and Introduction
• Wind Turbines: Dataset Download and Introduction
• Others:
  • Numenta's NAB
  • Yahoo's Webscope S5---Download Harder
  • 2020 Skoltech's SKAB
  • 2018 AIOps's KPI-Anomaly-Detection

4. Evaluate

4.1 Metrics

| Ground truth | Predict | Predict | |:-----------------:|:-----------------:|:-------------:| | | Abnormal | Normal | | Abnormal | TP | FN | | Normal | FP | TN |

• Precision: $P=\frac{TP}{TP+FP}$

• Recall: $R=\frac{TP}{TP+FN}$

• F1: $F1=\frac{2\times P\times R}{P+R}$

• AUC: $\mathrm{TPR}=\frac{TP}{TP+FN}$ $\mathrm{FPR}=\frac{FP}{TN+FP}$

4.2 Threshold

4.2.1 Label-Based Threshold Search

• Best F1

4.2.2 Thresholds Search without labels

• $Val_{max}(Train_{max})$ F1

3 sigma rule： $Val_{max}(Train_{max}) \approx mean + 3 \times std$

• Pot F1
• Epsilon F1

5. Point Adjust & Point Adjust %K & Original

• Point Adjust

<div style="text-align: center"> <img src="https://github.com/qiumiao30/time-series-anomaly-detection/blob/main/image/point%20adjust.png"/> </div>