Fair Preprocessing

This repository contains the source code and data used for the following paper, appeared at ESEC/FSE 2021. The repository been also evaluated and acceped in the artifact track of the conference. For any questions, contact the corresponding author. The replication package is licensed under MIT License: Copyright (c) 2020 Sumon Biswas

Title Fair Preprocessing: Towards Understanding Compositional Fairness of Data Transformers in Machine Learning Pipeline

Authors Sumon Biswas (sumon@iastate.edu) and Hridesh Rajan (hridesh@iastate.edu)

PDF https://arxiv.org/abs/2106.06054

Index

1. Benchmark
2. Installation and Evaluation
3. Datasets

• German Credit
• Adult Census
• Bank Marketing
• Compas
• Titanic

4. Source code

• Experiments

• Fair preprocessing
• Fair transformers

• Utilities

5. Results (RQ1, RQ2, RQ3)
6. DOI and Citation

Benchmark

The benchmark contains 37 ML pipelines under 5 different tasks from three prior studies.

German Credit	Adult Census	Bank Marketing	Compas	Titanic
GC1	AC1	BM1	CP1	TT1
GC2	AC2	BM2	-	TT2
GC3	AC3	BM3	-	TT3
GC4	AC4	BM4	-	TT4
GC5	AC5	BM5	-	TT5
GC6	AC6	BM6	-	TT6
GC7	AC7	BM7	-	TT7
GC8	AC8	BM8	-	TT8
GC9	AC9	-	-	-
GC10	AC10	-	-	-

DOI of Replication Package

Cite the paper as:

@inproceedings{biswas21fair,
  author = {Sumon Biswas and Hridesh Rajan},
  title = {Fair Preprocessing: Towards Understanding Compositional Fairness of Data Transformers in Machine Learning Pipeline},
  booktitle = {ESEC/FSE'2021: The 29th ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering},
  location = {Athens, Greece},
  year = {2021},
  entrysubtype = {conference},
  url = {https://doi.org/10.1145/3468264.3468536},
}