Please note that the project is still in beta phase. Please report any issues you encounter or suggestions you have. We will do our best to address them quickly. Contributions are also highly welcome!

NeuralProphet

A Neural Network based Time-Series model, inspired by Facebook Prophet and AR-Net, built on PyTorch.

Documentation

We are currently working on an improved documentation page.

For a visual introduction to NeuralProphet, view the presentation given at the 40th International Symposium on Forecasting.

Discussion and Help

Discuss with our community here on Github

Tutorials

There are several example notebooks to help you get started.

Please refer to our documentation page for more resources.

Minimal example

from neuralprophet import NeuralProphet

After importing the package, you can use NeuralProphet in your code:

m = NeuralProphet()
metrics = m.fit(df, freq="D")
future = m.make_future_dataframe(df, periods=30)
forecast = m.predict(future)

You can visualize your results with the inbuilt plotting functions:

fig_forecast = m.plot(forecast)
fig_components = m.plot_components(forecast)
fig_model = m.plot_parameters()

Install

You can now install neuralprophet directly with pip:

pip install neuralprophet

If you plan to use the package in a Jupyter notebook, we recommended to install the 'live' version:

pip install neuralprophet[live]

This will allow you to enable plot_live_loss in the fit function to get a live plot of train (and validation) loss.

If you would like the most up to date version, you can instead install direclty from github:

git clone <copied link from github>
cd neural_prophet
pip install .

Please note that NeuralProphet requires a Python version >= 3.7 due to the use of @dataclass.

Contribute

Dev Install

Before starting it's a good idea to first create and activate a new virtual environment:

python3 -m venv <path-to-new-env>
source <path-to-new-env>/bin/activate

Now you can install neuralprophet:

git clone <copied link from github>
cd neural_prophet
pip install -e .[dev]
neuralprophet_dev_setup
git config pull.ff only

Notes:

Including the optional -e flag will install neuralprophet in "editable" mode, meaning that instead of copying the files into your virtual environment, a symlink will be created to the files where they are.
The neuralprophet_dev_setup command runs the dev-setup script which installs appropriate git hooks for Black (pre-commit) and Unittests (pre-push).
setting git to fast-forward only prevents accidental merges when using git pull.

Style

We deploy Black, the uncompromising code formatter, so there is no need to worry about style. Beyond that, where reasonable, for example for docstrings, we follow the Google Python Style Guide

As for Git practices, please follow the steps described at Swiss Cheese for how to git-rebase-squash when working on a forked repo.

Changelogs

Current model features

Autocorrelation modelling through AR-Net
Piecewise linear trend
Fourier term Seasonality at different periods such as yearly, daily, weekly, hourly.
Lagged regressors
Future regressors
Holidays & special events
Sparsity of coefficients through regularization
Plotting for forecast components, model coefficients as well as final forecasts
Automatic selection of training related hyperparameters

Coming up soon

For details, please view the Development Timeline.

The next versions of NeuralProphet are expected to cover a set of new exciting features:

Robustify training for different datasets
Logistic growth for trend component.
Uncertainty estimation of individual forecast components as well as the final forecasts.
Support for panel data by building global forecasting models.
Incorporate time series featurization for improved forecast accuracy.

0.2.8 (next)

soft-start regularization
confidence interval for forecast (as quantiles via pinball loss)

0.2.7 (current)

example notebooks: Sub-daily data, Autoregresseion
bugfixes: lambda_delay, train_speed

0.2.6

Auto-set batch_size and epochs
add train_speed setting
add set_random_seed util
continued removal of AttrDict uses
bugfix to index issue in make_future_dataframe

0.2.5

documentation pages added
1cycle policy
learning rate range test
tutorial notebooks: trend, events
fixes to plotting, changepoints

Authors

The project efford is led by Oskar Triebe (Stanford University), advised by Nikolay Laptev (Facebook, Inc) and Ram Rajagopal (Stanford University) and has been partially funded by Total S.A. The project has been developed in close collaboration with Hansika Hewamalage, who is advised by Christoph Bergmeir (Monash University).

Contributors

This is the list of NeuralProphet's significant contributors. This does not necessarily list everyone who has contributed code. To see the full list of contributors, see the revision history in source control.

Oskar Triebe
Hansika Hewamalage
Nikolay Laptev
Riley Dehaan
Gonzague Henri
Ram Rajagopal
Christoph Bergmeir
Italo Lima
Caner Komurlu
Rodrigo Riveraca

If you are interested in joining the project, please feel free to reach out to me (Oskar) - you can find my email on the AR-Net Paper.

park-moonkyu/neural_prophet