Module detection 

Change-point and shift detection.

Algorithms for detecting process mean shifts and trend changes, covering both online (sequential surveillance) and offline (retrospective) approaches.

Online Detection (Sequential Surveillance)

• Cusum — Cumulative Sum chart (Page, 1954) for detecting small persistent shifts
• Ewma — Exponentially Weighted Moving Average chart (Roberts, 1959)

Offline Detection (Retrospective Changepoint Analysis)

• Pelt — Pruned Exact Linear Time algorithm (Killick et al., 2012) for detecting multiple changepoints with O(n) expected complexity
• SpectralResidual — Saliency-based one-shot anomaly scoring (Ren et al., 2019): spikes, steps and dropouts without a trained model

References

• Page, E.S. (1954). “Continuous Inspection Schemes”, Biometrika 41(1/2), pp. 100-115.
• Roberts, S.W. (1959). “Control Chart Tests Based on Geometric Moving Averages”, Technometrics 1(3), pp. 239-250.
• Killick, R., Fearnhead, P., & Eckley, I.A. (2012). “Optimal Detection of Changepoints with a Linear Computational Cost”, JASA 107(500), pp. 1590-1598.
• Ren, H. et al. (2019). “Time-Series Anomaly Detection Service at Microsoft”, KDD 2019, pp. 3009-3017.

Structs

Cusum

CUSUM chart parameters and state.

CusumResult

Result of CUSUM analysis for a single observation.

Ewma

EWMA chart parameters.

EwmaResult

Result of EWMA analysis for a single observation.

MultiPeltResult

Result of multivariate PELT changepoint detection.

Pelt

PELT changepoint detector.

PeltResult

Result of PELT changepoint detection.

SpectralResidual

Spectral residual anomaly scorer.

SrPoint

One scored point.

Enums

CostFunction

Cost function for evaluating segment homogeneity.

Penalty

Penalty selection for the PELT algorithm.

SpectralResidualError

Why a SpectralResidual run was refused.

Constants

SR_MIN_OBSERVATIONS

Fewest observations the transform is applied to at once.