Best Open‑Source Forecasting Software For Data Science Teams

Robust additive model for daily, weekly, and yearly seasonality with automatic changepoint detection; works in Python, R, and Stan.

Neural network extension of Prophet that adds auto‑regression, lagged regressors, and custom loss functions.

Amazon‑backed library for probabilistic time‑series modeling, featuring DeepAR, Transformer, and other state‑of‑the‑art models.

High‑level API built on PyTorch Lightning for rapid development of deep learning forecasting models (e.g., N‑BEATS, TFT).

Unified framework supporting classical, machine‑learning, and deep learning models with easy back‑testing and ensembling.

Facebook‑research toolkit offering a wide range of statistical and deep learning models, plus utilities for feature extraction and evaluation.

Python library providing classic econometric and statistical models (ARIMA, SARIMAX, VAR, etc.) with extensive diagnostics.

Scikit‑learn compatible toolbox for time‑series learning, covering forecasting, classification, and regression.

Automated model selection and hyper‑parameter tuning for dozens of forecasting algorithms.

LinkedIn’s forecasting library that combines statistical and machine‑learning methods with a focus on interpretability.

Low‑code machine‑learning library that automates preprocessing, model training, and deployment for time‑series data.

Bayesian time‑series modeling library offering state‑space models, GARCH, and ARIMA with MCMC inference.

Machine‑learning toolbox dedicated to time‑series clustering, classification, and regression with distance‑based methods.

Julia package providing a collection of statistical forecasting models and utilities for rapid experimentation.

Julia framework for time‑series manipulation, resampling, and integration with forecasting models.

Comprehensive R package for exponential smoothing, ARIMA, TBATS, and hierarchical forecasting.

Modern tidyverse‑compatible forecasting toolbox offering model building, evaluation, and ensemble methods.

Tidy temporal data structure that underpins the fable ecosystem, simplifying data wrangling for forecasts.

Deep neural architecture for univariate forecasting that achieves state‑of‑the‑art performance without external data.

Google‑research model that blends attention mechanisms with interpretable gating for multivariate forecasting.

TensorFlow library offering probabilistic forecasting layers, preprocessing utilities, and model evaluation tools.

Open‑source data processing engine that can run real‑time forecasting tasks on InfluxDB time‑series streams.

Visual programming add‑on for the Orange data‑mining suite, providing interactive forecasting widgets.

Collection of benchmark models and utilities used in the M5 competition, supporting both statistical and ML approaches.

High‑performance Python library (C++ backend) for statistical forecasting models like ETS, ARIMA, and Prophet.

Ready‑to‑use implementations of popular forecasting models (e.g., N‑BEATS, DeepAR) built on PyTorch Lightning.

R package for real‑time epidemiological forecasting, now part of the broader EpiForecasts ecosystem.

Probabilistic state‑space model implementation within GluonTS, useful for irregularly spaced series.

Auto‑regressive recurrent network for probabilistic forecasting, included as a core model in GluonTS.

R wrapper for Meta’s Prophet model, enabling seamless integration with tidyverse pipelines.

Example notebooks and utilities for applying TFT models to business forecasting tasks.

AutoML extension of Amazon’s AutoGluon focused on time‑series, automatically selecting models and hyper‑parameters.

Reference implementations of top‑performing models from the M5 competition, useful as starting points.

Specialized component of PyFlux for volatility modeling using GARCH and EGARCH families.

R package offering recurrent neural network models (LSTM, GRU) for multivariate time‑series forecasting.

Fork of Prophet that adds STL decomposition and custom seasonalities for more granular control.

Python library aggregating cutting‑edge deep learning forecasting models (N‑BEATS, DeepAR, TFT) with a unified API.