OAGRE : Outlier Attenuated Gradient Boosting Regression

Status: Functional - 

A meta-learning model for regression on noisy data with heteroscedasticity.

This work was initially started in 2017 while working with a large scale noisy regression problem. The initial experiments were done in R and abandoned when I moved onto other projects. The same line of thinking recurred in 2021 as I looked at more regression problems and led to this repository.

This time round the implementation has been done in Python in a scikit-learn compatible structure. It also allows you to define the internal classifier and regression algorithms to be used, rather than forcing the use of decision trees.

Massive thanks are required to the contributors at the scikit-lego project for the inspiring open-source library that informed much of the development here.

Installation

The package will be released via PyPi and can installed via pip.

Alternatively you can install from source code

Experiments

We have conducted experiments using synthetically generated data for highly non-linear regression problems and multiple variations of heteroscedastic noise. These experiments can be executed using the script run_experiment.py and the analysed with the script scripts/analyse.py.

How to cite

Paper to be published (under revision)

@InProceedings{Hawkins2024,
   author = {John Hawkins},
   year = {2024},
   title = {OAGRE: Outlier Attenuated Gradient Boosted Regression},
   booktitle = {},
   month = {},
}