Transforming Recruitment Systems Through Advanced Machine Learning for Bias Reduction and Operational Efficiency

This research presents a comprehensive machinelearning framework that enables recruitment decision-making by applying bias-mitigation techniques through the use of explainable AI. The framework is based on a large dataset with demographic and skill-based features, and multiple classifiers (e.g., Random Forest, XGBoost, and CatBoost) were utilized to build and tune the models. Interpretability techniques (SHAP and LIME) improved the interpretability of the models and demonstrated how key features affect hiring decisions while ensuring transparency. The results showed that all classifiers had high levels of predictive accuracy, with CatBoost being the highest performing. In addition, a comparative analysis of SHAP and LIME demonstrated how they provided complementary benefits (global vs. local/explanation vs. insight) to promote fairness and accountability in recruitment. Thus, this research provides an effective interpretative framework for automating hiring decisions while also addressing ethical concerns associated with the process. Further, this study provides a foundation for future research integrating innovative bias-mitigation techniques into machine-learning frameworks.