Early identification of students at risk of dropping out is a critical task in Learning Management Systems (LMS), enabling timely interventions that can improve academic outcomes and reduce attrition. While machine learning (ML) models have been widely applied for this purpose, achieving a balance between accuracy and interpretability remains challenging. This study proposes a hybrid ensemble learning framework that first uses an Artificial Neural Network (ANN) for extracting deep feature representations from student activity data. Then it applies classical ML classifiers to these learned features for dropout prediction. Evaluated on a real-world dataset covering four years of academic records for training and one year for testing, the proposed models were assessed using accuracy, precision, recall, and F1-score. Experimental results demonstrate that most hybrid models outperform their standalone counterparts, with notable gains in recall and F1-score for ANN-DT and ANN-NB, and the ANN-LSTM model achieving 98% accuracy. These results validate the framework’s effectiveness in enhancing predictive performance while maintaining simplicity and adaptability for practical LMS deployment.