Online Time Series Anomaly Detectors
Contextual Anomaly Detector - Open Source (CAD)
Classic processing KNN based Conformal Anomaly Detector (KNN-CAD)
Classic Processing Probabilistic-EWMA (PEWMA).
Classic Processing Shift-Detection based on EWMA (SD-EWMA).
Classic Processing Two-Stage Shift-Detection based on EWMA
Get detector score
Get Lables
Get Null And Perfect Scores
Get Number of Training Values
Get Window Length
Get windows limits
Incremental processing KNN based Conformal Anomaly Detector (KNN-CAD).
Incremental Processing Probabilistic-EWMA (PEWMA).
Incremental Processing Shift-Detection based on EWMA (SD-EWMA).
Incremental Processing Two-Stage Shift-Detection based on EWMA
Normalize Score using Max and Min normalization
Optimized Classic Processing Probabilistic-EWMA (PEWMA).
Optimized Classic Processing Shift-Detection based on EWMA (SD-EWMA).
Optimized Classic Processing Two-Stage Shift-Detection based on EWMA
Optimized Incremental Processing Probabilistic-EWMA (PEWMA).
Optimized Incremental Processing Shift-Detection based on EWMA (SD-EWM...
Optimized Incremental Processing Two-Stage Shift-Detection based on EW...
PLOT DETECTIONS
Reduce Anomalies
Implements a set of online fault detectors for time-series, called: PEWMA see M. Carter et al. (2012) <doi:10.1109/SSP.2012.6319708>, SD-EWMA and TSSD-EWMA see H. Raza et al. (2015) <doi:10.1016/j.patcog.2014.07.028>, KNN-CAD see E. Burnaev et al. (2016) <arXiv:1608.04585>, KNN-LDCD see V. Ishimtsev et al. (2017) <arXiv:1706.03412> and CAD-OSE see M. Smirnov (2018) <https://github.com/smirmik/CAD>. The first three algorithms belong to prediction-based techniques and the last three belong to window-based techniques. In addition, the SD-EWMA and PEWMA algorithms are algorithms designed to work in stationary environments, while the other four are algorithms designed to work in non-stationary environments.