Orbit | Uber's New Open Source Python Library for Time Series Modeling
Orbit
is Uber’s new python package for time series modeling and inference using Bayesian sampling methods for model estimation.

Orbit provides a familiar and intuitive initialize-fit-predict interface for working with time series tasks, while utilizing probabilistic modeling under the hood.
As per Orbit’s documentation, initial release supports concrete implementation for the following models:
- Local Global Trend (LGT)
- Damped Local Trend (DLT)
Both models, which are variants of exponential smoothing, support seasonality and exogenous (time-independent) features.
The initial release also supports the following sampling methods for model estimation:
- 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
Quick Start
Installation
pip install orbit-ml
Orbit requires PyStan as a system dependency. PyStan is licensed under GPLv3 , which is a free, copyleft license for software.
Data
iclaims_example
is a dataset containing the weekly initial claims for US unemployment benefits against a few related google trend queries from Jan 2010 - June 2018.
Number of claims are obtained from Federal Reserve Bank of St. Louis while google queries are obtained through Google Trends API.
Quick Starter Code
|
|
