Skip to contents

AI Fairness 360 (AIF360) R Package

Overview

The AI Fairness 360 toolkit is an open-source library to help detect and mitigate bias in machine learning models. The AI Fairness 360 R package includes a comprehensive set of metrics for datasets and models to test for biases, explanations for these metrics, and algorithms to mitigate bias in datasets and models.

Installation

Install the CRAN version:

Or install the development version from GitHub:

# install.packages("devtools")
devtools::install_github("Trusted-AI/AIF360/aif360/aif360-r")

Then, use the install_aif360() function to install AIF360:

Installation methods

AIF360 is distributed as a Python package and so needs to be installed within a Python environment on your system. By default, the install_aif360() function attempts to install AIF360 within an isolated Python environment (“r-reticulate”).

You can check using reticulate::conda_python() and reticulate::py_config()

Suggested steps

  1. Install reticulate and check if you have miniconda installed. If you do, go to step 2.
install.packages("reticulate")
reticulate::conda_list()

If you get an error: Error: Unable to find conda binary. Is Anaconda installed?, please install miniconda

reticulate::install_miniconda()

If everything worked, you should get the message:

* Miniconda has been successfully installed at '/home/rstudio/.local/share/r-miniconda'.

You can double check:

reticulate::conda_list()

You will get something like this:

          name                                                              python
1  r-miniconda                   /home/rstudio/.local/share/r-miniconda/bin/python
2 r-reticulate /home/rstudio/.local/share/r-miniconda/envs/r-reticulate/bin/python
  1. You can create a new conda env and then configure which version of Python to use:
reticulate::conda_create(envname = "r-test")
reticulate::use_miniconda(condaenv = "r-test", required = TRUE)

Check that everything is working reticulate::py_config().

  1. If you haven’t yet, please install the aif360 package install.packages("aif360") and then install aif360 dependencies
aif360::install_aif360(envname = "r-test")

Note that this step should take a few minutes and the R session will restart.

  1. You can now activate your Python environment
reticulate::use_miniconda(condaenv = "r-test", required = TRUE)

Getting Started 

# load a toy dataset
data <- data.frame("feature1" = c(0,0,1,1,1,1,0,1,1,0),
                   "feature2" = c(0,1,0,1,1,0,0,0,0,1),
                   "label" = c(1,0,0,1,0,0,1,0,1,1))

# format the dataset
formatted_dataset <- aif360::binary_label_dataset(data_path = data,
                                          favor_label = 0,
                                          unfavor_label = 1,
                                          unprivileged_protected_attribute = 0,
                                          privileged_protected_attribute = 1,
                                          target_column = "label",
                                          protected_attribute = "feature1")

Troubleshooting

If you encounter any errors during the installation process, look for your issue here and try the solutions.

Locked binding

If you get an error: cannot change value of locked binding, please restart the R session. Then try reactivating your Python environment and running the following commands exactly once:

Contributing

If you’d like to contribute to the development of aif360, please read these guidelines.

Please note that the aif360 project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.