Exploring Bias Mitigation Techniques for Predictive Analytics in College Student Success

This dissertation explores the pressing issue of algorithmic fairness and bias in machine learning (ML) models, with a particular focus on educational contexts. While ML models traditionally aim to maximize predictive accuracy, this research scrutinizes the fairness of predictions, especially concerning racially minoritized students in college success prediction models. The study emphasizes group fairness and investigates techniques to mitigate racial bias in predictive outcomes. Utilizing a nationally representative dataset, this research evaluates the effectiveness of various ML models and bias mitigation strategies. The findings contribute to the creation of a data science toolkit specifically designed for higher education institutions. This toolkit aims to enhance the accuracy and fairness of college success predictions, fostering more equitable academic outcomes for diverse demographic groups.