This project
- is stable and being incubated for long-term support. It may contain new experimental code, for which APIs are subject to change.
- requires PyStan as a system dependency. PyStan is licensed under GPLv3, which is a free, copyleft license for software.
Orbit is a Python package for Bayesian time series forecasting and inference. It provides a familiar and intuitive initialize-fit-predict interface for time series tasks, while utilizing probabilistic programming languages under the hood.
For details, check out our documentation and tutorials:
- HTML (stable): https://orbit-ml.readthedocs.io/en/stable/
- HTML (latest): https://orbit-ml.readthedocs.io/en/latest/
Currently, it supports concrete implementations for the following models:
- Exponential Smoothing (ETS)
- Local Global Trend (LGT)
- Damped Local Trend (DLT)
- Kernel Time-based Regression (KTR)
It also supports the following sampling/optimization methods for model estimation/inferences:
- Markov-Chain Monte Carlo (MCMC) as a full sampling method
- Maximum a Posteriori (MAP) as a point estimate method
- Variational Inference (VI) as a hybrid-sampling method on approximate distribution
Install from PyPi:
$ pip install orbit-ml
Install from source:
$ git clone https://github.com/uber/orbit.git
$ cd orbit
$ pip install -r requirements.txt
$ pip install .
$ pip install git+https://github.com/uber/orbit.git@dev
from orbit.utils.dataset import load_iclaims
from orbit.models import DLT
from orbit.diagnostics.plot import plot_predicted_data
# log-transformed data
df = load_iclaims()
# train-test split
test_size = 52
train_df = df[:-test_size]
test_df = df[-test_size:]
dlt = DLT(
response_col='claims', date_col='week',
regressor_col=['trend.unemploy', 'trend.filling', 'trend.job'],
seasonality=52,
)
dlt.fit(df=train_df)
# outcomes data frame
predicted_df = dlt.predict(df=test_df)
plot_predicted_data(
training_actual_df=train_df, predicted_df=predicted_df,
date_col=dlt.date_col, actual_col=dlt.response_col,
test_actual_df=test_df
)
Forecasting / Nowcasting with Regression in DLT (based on v1.0.13):
Backtest on M3 Data:
More examples can be found under tutorials and examples.
We welcome community contributors to the project. Before you start, please read our code of conduct and check out contributing guidelines first.
We document versions and changes in our changelog.
Check out the deck for a meet-up event in July 2021.
To cite Orbit in publications, refer to the following whitepaper:
Orbit: Probabilistic Forecast with Exponential Smoothing
Bibtex:
@misc{
ng2020orbit,
title={Orbit: Probabilistic Forecast with Exponential Smoothing},
author={Edwin Ng,
Zhishi Wang,
Huigang Chen,
Steve Yang,
Slawek Smyl},
year={2020}, eprint={2004.08492}, archivePrefix={arXiv}, primaryClass={stat.CO}
}
- Bingham, E., Chen, J. P., Jankowiak, M., Obermeyer, F., Pradhan, N., Karaletsos, T., Singh, R., Szerlip, P., Horsfall, P., and Goodman, N. D. Pyro: Deep universal probabilistic programming. The Journal of Machine Learning Research, 20(1):973โ978, 2019.
- Hoffman, M.D. and Gelman, A. The No-U-Turn sampler: adaptively setting path lengths in Hamiltonian Monte Carlo. J. Mach. Learn. Res., 15(1), pp.1593-1623, 2014.
- Hyndman, R., Koehler, A. B., Ord, J. K., and Snyder, R. D. Forecasting with exponential smoothing: the state space approach. Springer Science & Business Media, 2008.
- Smyl, S. Zhang, Q. Fitting and Extending Exponential Smoothing Models with Stan. International Symposium on Forecasting, 2015.