Nixtla/neuralforecast

Health v2 · maintenance only

Adoption not part of the score

4257 stars · 499 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

NeuralForecast is a Python library offering a large collection of state-of-the-art neural forecasting models (NBEATS, NHITS, TFT, PatchTST, DeepAR, LSTM, transformers, and more) built on PyTorch. It provides a unified sklearn-like fit/predict interface with support for exogenous variables, probabilistic forecasting, and automatic hyperparameter selection.

Use cases

• forecast future sales or demand with neural networks
• train NBEATS or NHITS on multiple time series
• generate probabilistic forecasts with quantile intervals
• forecast time series with exogenous variables like prices or weather
• automatically tune and select the best forecasting model
• benchmark deep learning models against statistical baselines

When to choose

• you need deep learning models for time series forecasting in Python
• you want a simple unified API across many neural forecasting architectures
• you need probabilistic or quantile forecasts
• you have many related time series and want global models with cross-learning

When to avoid

• you only need simple statistical methods like ARIMA or ETS
• you cannot use PyTorch or lack GPU resources for large-scale training
• you need a hosted foundation-model API rather than training your own models

Facets

library · maturity active

machine-learning deep-learning data-science sdk machine-learning deep-learning data-science time-series python cross-platform time-series forecasting pytorch nbeats nhits transformer probabilistic-forecasting exogenous-variables hyperparameter-tuning gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="Nixtla/neuralforecast")