MLForecast

MLForecast is an open-source Python library for data scientists and developers who need to forecast many time series with machine learning. It combines time series feature engineering with scikit-learn-compatible models, so teams can use regressors such as LightGBM and LinearRegression rather than adopt a dedicated forecasting model. Its license is Apache 2.0.

The library handles the work of turning historical observations into model inputs. It supports lagged values, rolling and expanding averages, and calendar features, along with transformations of the value being forecast. Models can learn across multiple series in one dataset, which suits applications such as retail demand and electricity load forecasting.

MLForecast works with pandas and Polars dataframes. For larger workloads, it supports distributed training through Dask, Ray and Spark, including remote clusters. That gives teams a choice between a Python workflow on their own machine and training across cluster infrastructure.

Forecasts can include prediction intervals through conformal prediction, rather than only a single expected value. External variables and fixed attributes of each series can also inform predictions. Cross-validation supports model evaluation and selection across multiple time series, while transfer learning lets a model trained on one set of series forecast another set.