Anomaly detection with PyOD/PyGOD

[cshjin]

Settings

• data dimension $\mathbb{R}^{W \times N \times F}$ with
  • $W$: number of workflows
  • $N$: number of nodes
  • $F$: number of features
• job-level anomaly detection
  • tabular data with selected features
    • single job data dimension $\mathbb{R}^{W \times F}$
    • single workflow data dimension $\mathbb{R}^{N \times F}$
    • all jobs data dimension $\mathbb{R}^{WN \times F}$
• workflow-level anomaly detection
  • any detected anomaly in job-level within the workflow leads to the anomaly in workflow-level
• collect the results as base cases for comparison

Reference:

• PyOd paper, code
• PyGOD paper, code