An Effective Online Learning Course Recommendation Using Improved Deep Active Convolutional Neural Network Based Sentiment Analysis and Ranking

Online learning platforms are used to discover the optimal learning courses for learners according to their interests and knowledge. An effective methodology is needed to suggest effective learning courses according to Sentiment Analysis (SA). It is difficult to handle large user feedback manually, so the recommendation system is utilized. The recommender system should be developed with high efficiency in filtering information. Also, it requires efficient access and resolves the issue of information overload. In order to solve this issue, effective deep learning-based approaches for online course ranking were presented in this paper. The input dataset used for this work contains the information on the online course. Initially, the input text data is pre-processed using different effective pre-processing approaches. Afterwards, features such as Improved Term Frequency-Inverse Document Frequency (ITF-IDF), Bag of Words (BoW), and glove word embedding are extracted to enhance the classification performance. Further, the Modified Rain Optimization (MRO) algorithm is utilized for feature selection by reducing the redundant features. Finally, an Improved Deep active Convolutional Neural Network (IDCNN) is presented for online course preference SA. Here, the Adaptive Beetle Antennae optimization algorithm (ABA) is utilized for weight optimization in the proposed IDCNN. This enhanced IDCNN predicts effective online courses by using SA as positive, negative, and neutral. Finally, the optimal learning course ranking is performed through the Jaccard similarity approach. This final recommendation through ranking improves the quality of the selection process for an online course. The presented methodology is implemented in the Python programming language. The experimental results proved that the presented approach attains enhanced performance on different performances like accuracy (98.17%), precision (98.23%), F1-score (98.19%), Root Mean Squared Error (RMSE) (0.21), Kappa (97.06%), recall (98.21%), and Area Under the Curve (AUC) (98.17%).