Meta's open-source Kats packs forecasting, anomaly detection and 65 time series features into one Python toolkit
mdancho84 · x · 2026-10-05
Data science educator mdancho84 spotlighted Kats, an open-source Python time series toolkit from Meta's Infrastructure Data Science team (6.5k GitHub stars), billed as a lightweight one-stop shop for time series analysis.
- Forecasting: integrates 10+ models including ARIMA, Prophet, Holt-Winters and VAR, plus ensembling, multivariate forecasts, backtesting, hyperparameter tuning, empirical prediction intervals and a self-supervised meta-learning model.
- Detection: changepoint, outlier/anomaly, trend and seasonality detection.
- TSFeatures: extracts 65 statistically defined time series features usable in downstream ML classification/regression.
- Utilities: includes time series simulators.
Tutorials ship as notebooks in the repo; the package is on PyPI.
More from coding & agent
- Same prompt, three tools: ChatGPT vs Claude vs Figma Make compared — Tegadesigns · 2026-10-05
- OntoPrune cuts 83% of context tokens, speeding local CPU TTFT by 6.7x — vigmarcarlo · 2026-10-05
- Debugging OpenCode stalls: Qwen3.6 hybrid memory forces llama-server full prompt re-processing — MysteriousInterest32 · 2026-10-05
- Veteran dev: AI cuts 3-4 week projects to 1-3 days, but engineering judgment still matters — moonean712 · 2026-10-05
- Managing a fleet of AI agents with a git repo of markdown plans and coordinator agents — bibryam · 2026-10-05
- Daniel Lemire proposes 'ephemeral testing': judge your library by building throwaway layers with AI — lemire · 2026-10-05