Explainable AI for Student Performance Prediction

Abstract

Student performance prediction is important in identification of poor performing students at the initial stages and offering early academic remedies to them. The black box the traditional models of predictions are based on traditional prediction models which provide and show correct results but the factors that affect the prediction are not explained. This does not play to the trust or informative decision-making of the educators. The paper hypothesizes a framework of Explainable Artificial Intelligence (XAI) with machine learning and deep learning algorithms consisting of Random Forest, XGBoost, and LSTM along with the tools of explainability, such as SHAP, LIME, and Attention Visualization. The system predicts students’ performance in terms of academic records, behavioral pattern, attendance and history of assessments as well as indicating how a specific prediction came to be. There sults of the experiments how that XAI increases the level of prediction reliability, transparency and acceptance among educators. It aids in making evidence- based decisions and assists institutions to offer individualized assistance, decreasing turnover, and increasing their educational results.