Sentiment-Based Hierarchical Deep Learning Framework Using Hybrid Optimization for Course Recommendation in E-learning

Abstract

Course recommendation (CD) is essential for success in a student’s educational journey. Due to the variations in student’s knowledge system, it might be difficult to select the course content from online educational platforms. This problem is overcome by introducing the Political Jellyfish search optimization (PJSO) based Hierarchical Deep Learning for Text (HDLTex) model for sentiment classification (SC) in CD. Here, the input data is taken from the E-khool database, which is subjected to the learner/course agglomerative matrix calculation. Then, the course is grouped by utilizing Bayesian Fuzzy clustering (BFC). When the query is given, bi-level matching is performed. The learner retrieves the preferred items after the best course group is found. Furthermore, course review data is applied to the tokenization process employing Bidirectional Encoder Representations from Transformers (BERT). Finally, the feature extraction is carried out and SC is performed by using HDLTex, which is trained by the proposed PJSO. Moreover, the PJSO is the incorporation of Political Optimizer (PO) and Jellyfish Search Optimization (JSO). The devised PJSO-based HDLTex has a superior assessment for maximum precision of 0.904, maximum recall of 0.915 and maximum F-Measure of 0.904 respectively.