Skip to content

Latest commit

 

History

History
32 lines (25 loc) · 844 Bytes

README.md

File metadata and controls

32 lines (25 loc) · 844 Bytes

DAMP

Source Code own
Learning type unsupervised
Input dimensionality multivariate

This approach uses the DAMP algorithm to find anomalies. Its functionality is described in this paper.

Output Format

The output will be an anomaly score for every subsequence of size anomaly_window_size.

Dependencies

  • python 3
  • numpy
  • pandas
  • stumpy

Notes

from timeeval.utils.window import ReverseWindowing
# post-processing for damp
def post_damp(scores: np.ndarray, args: dict) -> np.ndarray:
    window_size = args.get("hyper_params", {}).get("anomaly_window_size", 50)
    return ReverseWindowing(window_size=window_size).fit_transform(scores)