Forecasting The Adoption Rate Of The E Learning Using Multilayer Recurrent Neural Network With Long Short Term Memory On Analysis Of User Sentiment

Abstract

E learning has become a necessary option to the education system to the entire world due to occurrence of the covid -19 pandemic. Especially implementation of the lockdown, educational institution and students adapted to e-learning system. In order to evaluate the opinion of the students to e- learning system with respect to various challenges and advantages, sentiment analysis has been employed is used to gain valuable insight. The social networks are widely distributed to gather and share the user perspective. This textural information is highly sourced with the data providing the feelings of the students with the statements that expresses agreement or disagreement in the comment sections to reveal the negative or positive feelings of the students towards the learning for performing the sentiment analysis and opinion mining. With current technology advancement in the information technology, it is necessary to forecast the adaption rate of the e-learning system in future. Therefore, a new optimized Multilayer recurrent neural network with long short term memory architecture is proposed in this work to provide valuable forecasting and suggestion on the user intension. Model is adaptable to temporal and time varying data along its existence of long term dependences between the users. Initially data preprocessing is carried out user profiling. Preprocessed data is employed for Latent Dirichlet Allocation feature to extract the opinion on the latent user intention process on the parsed data and store long dependency of the data in latent representations on the LTSM Model. Latent user intention is projected to the long short term memory for efficient data organizing on indexing as network. Organized data is employed to Multilayer Recurrent Network composed of various layer is considered as learning representative which provides valuable forecasting as suggestion and recommendation on various aspects on user adoption. We evaluate the performance of the proposed deep learning approach on a twitter dataset which considered of 38602 tweets during covid -19 pandemic. It has been proved to be outperforming against state-of-the-art methods against precision, recall and F Measure respectively.