Interpretable Prediction of Student Dropout Using Explainable AI Models

Abstract

The research addresses the challenge of student dropouts in education institutions, recognizing the pivotal role of education in shaping the society. A research gap is identified, with limited studies effectively combining Machine Learning (ML) algorithms and Explainable AI (XAI) techniques for predicting student dropout. To bridge this gap, a prediction model is developed, integrating ML with XAI to enhance transparency and interpretability. A web interface is introduced, enabling users to input attributes and receive predictions with explanations interpreted using LIME. Further, global interpretation using SHAP is conducted, offering insights into the overall impact and importance of features in the ML model across its dataset. The model is evaluated by selecting academic professionals through purposive sampling technique, employing Likert scale questions, resulting in a mean value exceeding 3.4 and affirming the model’s effectiveness in predicting student dropouts with transparency.

Ähnliche Arbeiten