pyforest

Overview | Installation | Usage | Benchmarks | To-Do | Acknowledgements | License

Regression and classification with random forests in Stata

version 0.11 16jul2019

Overview

pyforest is an implementation of the random forest algorithm in Stata 16 for classification and regression. It is essentially a wrapper around the popular scikit-learn library in Python, making use of the Stata Function Interface to pass data to and from Python from within the Stata window.

Prequisites

pyforest requires Stata version 16 or higher, since it relies on the Python integration introduced in Stata 16.0. It also requires Python 3.x and the scikit-learn library. If you have not installed Python or scikit-learn, I would highly recommend starting with the Anaconda distribution.

Installation

There are two options for installing pyforest.

1. The most recent version can be installed from Github with the following Stata command:

net install pyforest, from(https://raw.githubusercontent.com/mdroste/stata-pyforest/master/)

2. A ZIP containing pyforest.ado and pyforest.sthlp can be downloaded from Github and manually placed on the user's adopath.

Usage

Basic usage of pyforest is pretty simple. The syntax looks similar to -regress-. Optional arguments share exactly the same syntax as scikit-learn.ensemble.RandomForestClassifier and scikit-learn.ensemble.RandomForestRegressor.

Here is a quick example demonstrating how to use pyforest for classification:

* load dataset of flowers
use http://www.stata-press.com/data/r10/iris.dta, clear

* mark approx half of the dataset for estimation
gen train = runiform()<0.5

* run random forest classification, save predictions as predicted_iris
pyforest iris seplen sepwid petlen petwid, type(classify) training_identifier(train) save_prediction(predicted_iris)

Here is a quick example demonstrating how to use pyforest for regression:

* load dataset of cars
sysuse auto, clear

* mark approx 30% of obs for estimation
gen train = runiform()<0.3

* run random forest regression, save predictions as predicted_price
pyforest price mpg trunk weight, type(regress) training_identifier(train) save_prediction(predicted_price)

(Incomplete) internal documentation can be found within Stata. This documentation is still a work in progress:

help pyforest

Finally, since the option syntax in this package is inherited from scikit-learn, the documentation for the scikit methods sklearn.ensemble.randomForestClassification and sklearn.ensemble.randomForestRegression may be useful.

Todo

The following items will be addressed soon:

• Finish off this readme.md and the help file
• Proide some benchmarking
• Make exception handling more robust
• Add support for weights
• Return some stuff in e()
• Post-estimation: feature importance
• Model selection: cross-validation

Acknowledgements

This program relies on the wonderful Python package scikit-learn.

License

pyforest is MIT-licensed.