Leveraging Machine Learning to Forecast Candidate Selection Outcomes

Machine learning has emerged as a transformative tool in education, driving personalized learning experiences. This study focuses on its application in the educational sector, particularly through the lens of peer learning systems. Our research presents a systematic approach to predicting candidate success during the selection phase, also known by the immersive evaluation phase. In this context, five distinct machine-learning algorithms (Decision Trees, Support Vector Machines, Logistic Regression, Random Forest, and Stacking) were employed to assess their effectiveness in classifying candidates as retained or rejected. Additionally, we explored attribute importance to provide insights into the key factors influencing candidate selection. The findings reveal that Stacking (Random Forest (RF), K-Nearest Neighbors (KNN), Naive Bayes (NB) + XGboost) model proved to be the most effective after evaluating performance metrics using bootstrapping methods.