Paper Type

Complete

Abstract

This study investigates whether data balancing techniques improve fairness across gender groups in stroke prediction using Logistic Regression on a highly imbalanced Kaggle stroke dataset. Six resampling methods were evaluated for predictive performance and gender fairness. Results show that resampling significantly enhances minority class detection, but resampling methods introduce significant gender disparities, indicating that data balancing does not inherently ensure fairness. These findings underscore the need for fairness-aware strategies in healthcare AI to mitigate biases while improving predictive accuracy.

Paper Number

1585

Comments

SIG DSA

Share

COinS
 
Aug 15th, 12:00 AM

Does Data Balancing Improve Fairness Across Gender Groups? A Comparative Study of Highly Imbalanced Stroke Prediction with a Logistic Regression Model

This study investigates whether data balancing techniques improve fairness across gender groups in stroke prediction using Logistic Regression on a highly imbalanced Kaggle stroke dataset. Six resampling methods were evaluated for predictive performance and gender fairness. Results show that resampling significantly enhances minority class detection, but resampling methods introduce significant gender disparities, indicating that data balancing does not inherently ensure fairness. These findings underscore the need for fairness-aware strategies in healthcare AI to mitigate biases while improving predictive accuracy.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.