Digital Twin-Enabled Personalized E-Learning Using Deep Learning and Real-Time Learning Analytics

Abstract

The rise of online learning platforms has led to the need for smart and personalized educational solutions. Existing e-learning systems offer standardized learning solutions that do not consider individual differences in learning preferences, styles, engagement levels, and other factors. This study proposes the Digital Twin-Enabled Personalized E-Learning Framework that combines deep learning and real-time learning analytics. This framework develops dynamic learner digital twins based on data provided by learning management systems, assessments, discussion forums, learning videos, and clickstream activities. The hybrid LSTM-DNN algorithm is used to predict learner performance, engagement levels, chances of completing a particular course, and even risk of dropping out of courses. Real-time learning analytics provides continuous updating of learners' profiles and personalized recommendations and adaptation. The proposed framework was tested using learners' interaction data in online learning environments. As a result of experimentation, the hybrid LSTM-DNN model was able to achieve a prediction accuracy of 95.2%, which is better than the prediction accuracy of conventional deep learning algorithms. The application of personalized learning paths led to a significant increase in learners' engagement levels, academic performance, resource usage, and course completion rates. Besides, the recommendation engine provided 93.8% recommendation accuracy. The results show that a combination of digital twin technology, deep learning, and real-time learning analytics can significantly improve personalization, retention rates, academic performance, and educational effectiveness of e-learning environments.