Imbalanced Data Classification Using Oversampling and Automatic Feature Selection Methods for Undergraduate Student Career Prediction.

The application of machine learning techniques for predicting the career trajectories of fresh undergraduate students has become a crucial strategy for evaluating their potential to secure employment post-graduation or pursue further education. However, for such applications, imbalanced data is a vital issue that needs to be addressed with proper methods. In this paper, the combination of oversampling, using Synthetic Minority Overs amp ling Technique (SMOTE) and Adaptive Synthetic Sampling (ADASYN), and feature selection, using Recursive Feature Elimination (RFE) and the Boruta algorithm, is applied. The results show that the SMOTE-based Boruta approach is effective to improve the performance of machine learning classification models for undergraduate student career prediction.