Gripa is a python package for time series anomaly detection. The name of Gripa is taken from Scandinavian languages, which means grab or catch or seize, and not meant as Flu in Spanish 😄.
- Gripa is light-weight. Try Gripa before you build LSTM or Transformer for your time series anomaly detection.
- Gripa is accurate. Gripa has two available algorithms that perform very well in detecting anomalies on your time series data.
- Gripa is comprehensive. Gripa can detect three types of anomaly: global anomaly, contextual anomaly, and level-shift.
Python 3.9or higher is required.
python -m pip install gripa# Load data
from gripa import Gripa
detector = Gripa()
anomalies = detector.fit_predict(data)There is only one module, gripa.Gripa, that can be used for detecting anomalies in time series.
class gripa.Gripa(window_size=11, algorithm="hpf", threshold=3)
window_size: int or float, default=11algorithm: {"hpf", "ssa"}, default="hpf"threshold: float, default=3
anomaly_score: score for labelling anomaly
fit(X): train the Gripa model, which can generate attributeanomaly_scorefit_predict(X): train and generate anomaly labels (TrueorFalse)
Hamid Dimyati - [email protected]