---
id: neuralprophet
version: "0.9.0"
license: MIT
license_treatment: permissive
maintenance: dormant
---

# neuralprophet — NeuralProphet is an easy to learn framework for interpretable time series forecasting.

License: permissive · Maintenance: dormant · Downloads: 96.4K/mo

## What it is and what it does

NeuralProphet is a time series forecasting framework built on PyTorch that blends neural networks with classical time-series methods. It targets developers who want to build, interpret, and iteratively refine forecasting models with minimal boilerplate. The framework includes components for autoregression, trend detection with changepoint identification, seasonality via Fourier terms, lagged and future regressors, holiday events, and quantile regression for uncertainty estimation. It is designed for sub-daily, multi-year datasets and emphasizes interpretability over out-of-the-box accuracy—the expectation is that you will adjust and iterate on results.

The package depends on numpy, pandas, torch, pytorch-lightning, tensorboard, torchmetrics, and visualization libraries (matplotlib, plotly, kaleido). It supports multiple time series with shared or local model parameters and includes built-in plotting for forecasts, components, and model coefficients. The project is in beta and dormant (last release June 2024), so while the codebase is stable and community-supported, you should expect limited active development and may need to handle compatibility issues with newer PyTorch versions yourself.

Use it for:

- Build sub-daily forecasting models (e.g., hourly energy demand, sensor readings) with interpretable trend and seasonality components.
- Fit a single global model across multiple related time series with partially shared parameters.
- Estimate forecast uncertainty by training quantile regression models for specific percentiles.
- Prototype and iterate on forecasting models interactively, inspecting component contributions and adjusting hyperparameters.
- Incorporate known future events (holidays, marketing campaigns) and lagged observations as model features.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

NeuralProphet is a PyTorch-based framework for interpretable time series forecasting that combines neural networks with traditional algorithms, designed for iterative model building and human-in-the-loop refinement.

Yes, if you need interpretable time series forecasting with iterative refinement and don't mind the dormant maintenance status. The low install friction, permissive MIT license, and active community support make it a reasonable choice for exploratory forecasting work. However, be aware that the last release was over 784 days ago and the project is in beta—if you require active bug fixes or compatibility guarantees with the latest PyTorch versions, consider alternatives or plan to contribute fixes yourself.

## Install

pip install neuralprophet
uv add neuralprophet
poetry add neuralprophet

## Installing neuralprophet

Before you install: Low install friction with a pure Python wheel. The package depends on 12 runtime libraries including torch, pytorch-lightning, and pandas. Maintenance is dormant—last release was 2024-06-21, over 784 days ago, though the repository remains active with recent commits and 4295 stars.

License in practice: MIT license is permissive, allowing commercial and private use with minimal restrictions. You may use, modify, and distribute the package freely as long as you include the license notice.

Quickstart:

pip install neuralprophet

from neuralprophet import NeuralProphet

m = NeuralProphet()
metrics = m.fit(df)
forecast = m.predict(df)
fig_forecast = m.plot(forecast)

Requires Python 3.9–3.12. Torch and pytorch-lightning are heavy dependencies; installation may take time depending on your environment. Data should be higher-frequency (sub-daily) and span at least two full periods/years for best results.

Verify before relying:

- Whether the 'dormant' maintenance status (784 days since last release) affects stability or compatibility with current PyTorch/Lightning versions.
- Real-world accuracy and performance compared to other forecasting frameworks on typical datasets.
- Whether the beta status (Development Status :: 4 - Beta) implies breaking changes in future releases.

## Package facts

- License: MIT (permissive)
- Python support: capped_below_current
- Install friction: low
- Maintenance: dormant
- Downloads: 96.4K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

time series forecasting, neural network forecasting, interpretable forecasting, pytorch time series, autoregressive forecasting, trend and seasonality detection, multivariate time series, time-series, forecasting, interpretability

[View on SkillFed](https://skillfed.io/packages/neuralprophet) · [View on PyPI](https://pypi.org/project/neuralprophet/)