Abstract
The abundance of data in educational databases makes it more difficult to predict students' success. Tutoring in the classroom and at home offers students individualised support and constructive criticism of their learning. Methods that correctly predict a student's performance in a particular cognitive analysis or course can help identify students who are at risk of failing a course during a pandemic and enable their educational institution to take appropriate action. These days, combining classifiers is proposed as a novel approach for enhancing prediction performance. The Ensemble Multiple Recurrent Deep Learning (EMRDL) algorithm has been put forth in this paper as a way to forecast student success under supportive learning conditions during the Coronavirus Disease (COVID-19) epidemic, which includes tutoring from families and schools. First, benchmark repository records are gathered. In order to select the most pertinent features and reduce the dataspace, the Circle Map Tuna-Swarm Optimization Algorithm (CMTSOA) is developed. A more precise feature search is made possible by the TSOA, circle map function, which has been added for random number generation. Following that, the EMRDL classifier is introduced to improve the categorization outcomes by fusing several models, including Ensemble Deep Long Short-Term Memory (EDLSTM), Weight Recurrent Multilayer Perceptron Network (WRMLP), and Gated Recurrent Unit (GRU). MLP has been introduced by adding delayed connections among nearby nodes of a hidden layer, an RMLP network is able to be built. Through connections established through a series of nodes, GRU is intended to conduct student performance prediction. By using majority voting, various classifiers are merged. By integrating the proposed system in Matrix Laboratory R 2020a (MATLAB R2020a) and using three distinct benchmark databases, the results of the proposed system and state-of-the-art classifiers is assessed (School, University, and C-19GA20). Metrics such as Precision, Recall, Accuracy, and Area Under Curve (AUC) are used to evaluate the performance of these models.
Keyword
Education data mining, Virtual learning, Circle Map Tuna-Swarm Optimization Algorithm (CMTSOA), Deep Learning, Ensemble Multiple Recurrent Deep Learning (EMRDL), Ensemble Deep Long Short-Term Memory (EDLSTM), Weight Recurrent Multilayer Perceptron Network (WRMLP), and Gated Recurrent Unit (GRU).
PDF Download (click here)
|