stldecompose 0.0.1

A Python implementation of seasonal trend with Loess (STL) time series decomposition

This is a relatively naive Python implementation of the “Seasonal and Trend decomposition using Loess” time series decomposition (“STL decomposition,” Cleveland et al. 1990 [pdf]).

This implementation is a variation of (and takes inspiration from) the current implementation of the seasonal_decompose method in statsmodels. In this implementation, the trend component is calculated by substituting a configurable Loess regression for the convolutional method used in seasonal_decompose. It also extends the existing DecomposeResult from statsmodels to allow for forecasting based on the calculated decomposition.

Usage

The stldecompose package is relatively lightweight, exposing only a couple of primary methods: decompose() and forecast(), along with a handful of built-in forecasting functions. See the included IPython notebook for more details.

Installation

A Python 3 virtual environment is recommended.

Current installation is via cloning this repo to your local workspace and a local pip install:

(env) $ git clone [e-mail]:jrmontag/STLDecompose.git
(env) $ cd STLDecompose; pip install .

More Resources

• statsmodels Time Series analysis package

• Hyndman’s OTexts reference on STL decomposition