An optimized deep learning framework based on LEE for real time student performance prediction in educational data

Abstract

Predicting student performance in real-time remains a critical challenge in educational data mining (EDM), especially with large, noisy, and high-dimensional datasets. This study proposes an advanced deep learning framework that integrates learning entropy estimation (LEE) with models such as support vector machines (SVM), you only look once (YOLO), recurrent convolutional neural networks (RCNN), and artificial neural networks (ANN) to enhance feature selection and classification accuracy. The framework follows a systematic pipeline involving data preprocessing, LEE-based feature extraction, and model training on a real-time academic dataset comprising student demographics, attendance, and performance metrics. Among the proposed models, the LEE-based YOLO (LBYOLO) achieved the highest testing accuracy of 93% and the fastest execution time of 1.84 seconds, while the LEE-based ANN (LBANN) demonstrated consistent performance across precision, recall, and F1-score. The results confirm the superiority of deep learning methods over traditional machine learning techniques for educational prediction tasks. This approach enables early detection of at-risk students and supports timely, data-driven educational interventions. Future work will focus on adaptive learning systems and multi-platform student behavior analysis to support personalized education strategies.