sktime: time series anomaly detection, changepoint detection, segmentation
summary
skchange extends sktime to detect point changes, changepoints and anomalous segments using concrete components: moving window, PELT, L2 and Gaussian cost classes compiled with numba, and a segment detector called kappa plus MV subset detection that identifies which variables are affected. detectors fit/predict return pandas integer iloc or pandas Interval outputs and transform yields per-row labels and scores. a HVAC case study extracts on/off weekly schedules successfully; a diabetes benchmark uses an FDA‑grade simulator with 337 meal events (~3.75/day) for a semi‑supervised meal detection challenge.
AI-written summary · about
chapters (9)
- 0:00Introduction and speakers
- 1:51What is sktime and SKchange
- 10:01Unified interface for time series
- 16:36Detection tasks overview
- 26:33Change detection with moving window
- 37:51Cost-based vs change score detectors
- 50:43Segment anomaly detection
- 1:05:13HVAC system use case
- 1:15:43Diabetes meal detection challenge
related talks
28mOnline Change Point Detection Using Spark Streaming
01h 28msktime - python toolbox for time series: advanced forecasting - probabilistic, glob...
45mAnomaly Detection in Time Series
01h 26mSktime—Python Toolbox for Time Series: How to Implement Your Own Estimator
30msktime: python toolbox for time series- next-generation AI
01h 30msktime - python toolbox for time series
32mAdopting Open-Source Tools for Time Series Forecasting
41mApplying Foundational Models for Time Series Anomaly Detection