A Cognitive–Generative Decision Intelligence Framework for Explainable Student Performance Prediction

Abstract

Making accurate predictions of students’ academic performance is a critical requirement in contemporary educational analytics, which helps in the early detection of at-risk students. Although machine learning models have shown remarkable predictive capabilities, they are not easily interpretable, which is less useful in real-world educational applications, where educators need not only predictions but also explanations to support those predictions. This paper proposes a new Cognitive-Generative Explainable AI Framework that combines predictive analytics with human-understandable explanations of decisions. The cognitive part of the framework uses ensemble techniques, combining Logistic Regression and Random Forest classifiers to make predictions of students’ academic performance based on academic and demographic variables. The generative part of the framework uses rule-based natural language generation techniques to produce structured explanations that are friendly to educators and aligned with the predictions made by the framework. The framework was tested on a balanced dataset of 1,430 student records, showing outstanding performance metrics of 98.9% accuracy, 98% precision, 99% recall, and 99% F1-score. Most importantly, the framework produces well-formed explanations that are aligned with predictions, which helps educators understand the outcome of the classification task and point to specific academic variables that influence students’ performance. This research work fills an important void in educational data mining by providing a comprehensive solution that balances high prediction accuracy with interpretability and explainability, thus ensuring trust, fairness, and usefulness in AI-based educational decision support systems.