Course Recommendation Using Hybrid Collaborative and Content Based Filtering Method Based on Optimized Roberta and Deep Learning Models in E-Learning Platform

E-learning has become a popular choice for learners during the pandemic, with many courses available on platforms. Recommender systems help users select the best courses based on their preferences, but the abundance of options and redundant information can make it difficult to find relevant courses. Existing recommendation models often struggle with insufficient or unrelevant data, leading to inaccurate predictions and poor decision-making. To overcome that, the proposed model introduced a hybrid collaborative and content based filtering method for course recommendation based on deep learning algorithm. First, the course data is gathered and preprocessed the collaborative features such as course name, difficulty, skills, and ratings, by removing stopwords, punctuation, and applying lemmatization and stemming to clean the text. Content features, like course descriptions, undergo preprocessing to eliminate and simplify words to their root form. After this, the data is tokenized and transformed into vector representations using the ROBERT a model, optimized with the Eel and Grouper Optimizer algorithms to fine-tune hyperparameters like learning rate and batch size. The model is then trained with a hybrid Bi-LSTM model, integrated with a 50-layer Deep Neural Network (DNN-50), to recommend courses effectively. The proposed approach achieves an accuracy of 98%, selectivity of 95.8%, and an F1 score of 96.5%. It outperforms existing models in providing accurate course recommendations. This makes it a more efficient solution for learners on e-learning platforms.