Deteksi Bias dalam Model Machine Learning untuk Prediksi Kelulusan Mahasiswa Berdasarkan Aktivitas Virtual Learning Environment

The rapid digital revolution in education has positioned Virtual Learning Environments (VLEs) as critical to evolving learning paradigms, particularly highlighted during the COVID-19 pandemic. This research investigates the impact of VLE-based student activity on predicting academic success and addresses the biases in machine learning models used for these predictions. Using the Open University Learning Analytics Dataset (OULAD), this study integrates data preprocessing techniques, feature selection, and data transformation to develop a comprehensive dataset. A Random Forest model is employed to predict student graduation outcomes, categorized into "pass", "fail", and "distinction" classes. The model's performance is evaluated using classification metrics such as accuracy, precision, recall, and F1-score, alongside confusion matrices. Bias detection is conducted using the DALEX tool, focusing on protected attributes like age, gender, and disability to ensure fairness. The results reveal high model accuracy but highlight significant bias in some demographic groups. This study contributes to the ongoing discourse on ensuring ethical and fair machine learning applications in educational settings by proposing methods to enhance the equity and transparency of predictive models.