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.
The output will be an anomaly score for every subsequence of size anomaly_window_size.
- python 3
- numpy
- pandas
- stumpy
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)