• Navjot Singh
• Shanu Khare
• Geetangali
• Payal Thakur

Bias in machine learning models can lead to unfair and discriminatory decision-making, affecting individuals and communities in various domains such as finance, healthcare, and criminal justice. To address this issue, researchers have proposed several fairness metrics and algorithms aimed at reducing bias and promoting unbiased decision-making. In this paper, we investigate the effectiveness of four fairness metrics and three algorithms in improving the fairness of machine learning models. Specifically, we consider demographic parity, equal opportunity, equality of odds, average odds difference, Reweighing, Disparate Impact Remover, and Prejudice Remover. We train logistic regression models on two real-world datasets, Adult Income and German Credit, and compare the fairness of the original and modified models using the above metrics and algorithms. Our results show that all four fairness metrics identify significant disparities in the original models, indicating the presence of bias. Moreover, Reweighing and Disparate Impact Remover improve demographic parity but worsen other metrics, suggesting a trade- off between different forms of fairness. On the other hand, Preju- dice Remover achieves the best overall performance in balancing multiple fairness metrics simultaneously, reducing the average odds difference by up to 70. Our findings highlight the need to carefully choose the appropriate fairness metric and algorithm depending on the specific context and application. Furthermore, transparency and accountability mechanisms should be put in place to monitor and mitigate potential biases in AI systems.

Machine Learning Bias, Fairness Metrics, Algorithms, Unbiased Decision Making, Demographic Parity, Equal Opportunity, Equality of Odds, Average Odds Difference, Reweighing, Disparate Impact Remover.

"Causal Discovery Using Graphical Models and Machine Learning Techniques", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.11, Issue 9, page no.d725-d729, September-2024, Available :http://www.jetir.org/papers/JETIR2409383.pdf

"Causal Discovery Using Graphical Models and Machine Learning Techniques", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.11, Issue 9, page no. ppd725-d729, September-2024, Available at : http://www.jetir.org/papers/JETIR2409383.pdf