Scribd
Upload
6 views4 pages

Feature Engineering Impact on Fairness

This document summarizes the results of preprocessing and feature engineering on four datasets - Adult Census, Bank Marketing, Home Credit, and Titanic. It shows the effect of imputing missing values, feature engineering causing bias, and developing semantics-free features to improve fairness. The key results are preprocessing reduced missing value proportions, feature engineering introduced bias with certain features showing higher effects, and developing semantics-free features improved fairness.

Uploaded by

Nabila Hassan
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as XLSX, PDF, TXT or read online on Scribd
6 views4 pages

Feature Engineering Impact on Fairness

This document summarizes the results of preprocessing and feature engineering on four datasets - Adult Census, Bank Marketing, Home Credit, and Titanic. It shows the effect of imputing missing values, feature engineering causing bias, and developing semantics-free features to improve fairness. The key results are preprocessing reduced missing value proportions, feature engineering introduced bias with certain features showing higher effects, and developing semantics-free features improved fairness.

Uploaded by

Nabila Hassan
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as XLSX, PDF, TXT or read online on Scribd

German Credit

SPD
EOD
AOD
ERD

Preprocessing
Drop off MV
SPD -0.05

0.09
EOD 0.09
AOD -0.1

0.04
ERD -0.3

0.02
1

-0.05

-0.1

-0.3
Adult Census Bank Marketing Home Credit Titanic

Preprocessing Stage- MV handling & Fairness


Impute(N/A) MV Impute (Mean)SPD
MV EOD Impute
AOD ERD
(Mean+Mode) MV
0.02 0.004 0.0001
0.09

0.04 0.002 0.0002


0.06

0.06 0.0001 0.0005


0.04

-0.05 -0.01 -0.001


0.02

0.0005
0.0001

0.0001

0.0002
0.004

0.002

1 2 3 4

-0.001
-0.01
-0.05

-0.05
-0.1

-0.3
FE causing Bias
SPD 0.5
Effect of feature engineer
EOD 0.76
AOD 0.8
0.8
ERD 0.05
0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
SPD EOD AO

Features causes bias Semantics


FE -fairness
of 0.005
feature engineering on fairness
0.001
0.004
0.0001

EOD AOD ERD

Features causes bias Semantics free features

You might also like