Abstract—Higher education institutions are focusing increasingly on targeted interventions by supporting students with academic challenges, improving their overall learning outcomes. In this context, hybrid Deep Learning (DL) approaches have emerged as leaders in the construction of recommendation systems capable of predicting students-at-risk by analyzing patterns of historical academic data. This paper proposes a hybrid DL model integrating an attention mechanism, specifically tailored to predict the success of students and to identify students-in-need of academic support. Our primary dataset, sourced from the Majmaah University of the Kingdom of Saudi Arabia, contained 146,989 records and 26 features, including a wide variety of student-related features crucial for making academic achievement predictions. Comparison with the current models like Artificial Neural Network (ANN), Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU), Neural Network Random Weight (NNRW), Long Short-Term Memory (LSTM), and NNRW-LSTM demonstrates the exemplary effectiveness of the proposed NNRW-LSTM model integrated with an attention mechanism. The proposed model achieved remarkable accuracy of 93.53%, 92.85% of the recall rate, 92.6% of precision, and 92.55 of the F1-Score. These results document the prospect of this hybrid DL model integrated with an attention mechanism to assist schools in identifying students-at-risk proactively, facilitating interventions well in time and helping curb academic underperformance and dropouts.
 
Keywords—deep learning, student success, higher study, Neural Network Random Weight-Long Short-Term Memory (NNRW-LSTM), attention mechanism


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).