A Deep Learning Framework for Adaptive E-Learning: Integrating Learning Style Detection in Web-Based Platforms

Modern digital learning has made the creation of the adaptive e-learning systems a priority because different needs of learners require individual approaches. The conventional approaches usually do not consider the learning styles of individual students, therefore, leading to less engagement and unequal academic results. To solve this, a deep learning-based model is proposed, which combines learning style identification with student performance forecasting with the Student Feedback Dataset of Kaggle. The framework uses sophisticated text preprocessing, extraction of features based on Word2Vec, as well as the classification based on LSTM model that is optimized on sequential data. The LSTM has excellent performance on experimental evaluation, reaching 99.29% accuracy (ACC), 94% precision (PRE), 95% recall (REC), and a 95% F1-score (F1) and validation loss of 0.2-0.4. Relative analysis indicates that LSTM is much better than SVM (75.55% ACC), BPNN (78.83), and ANN (79.11%). The framework is more inclusive, knowledge-retaining, and motivating by dynamically adjusting course delivery to personal learning preferences. Moreover, it provides teachers with practical knowledge to formulate powerful learning plans. Deep learning can transform online learning environments to be smarter, more customized and larger. This could be translated to better performance in school in the long-term.