darts

Why we included this project

For teams that need to forecast time series and also flag anomalies, Darts keeps both jobs in one library. It wraps dozens of models, from classic ARIMA to deep networks like N-BEATS, behind a consistent fit() and predict() interface, so you can swap baselines without rewriting code. The library also handles the practical workflow around forecasting: backtesting, blending multiple models, adding external covariates, and producing probabilistic forecasts with uncertainty ranges. The anomaly detection side is equally hands-on, letting you score series with PyOD-based detectors or wrap any forecasting model to spot unusual behavior in sensor or operational data.