Enhancing Learner's Performance In E-Learning Through Intelligent Techniques

Abstract

Smart education's adaptive e-learning solutions adjust the learning process to meet the specific requirements of each student. The crucial challenges such as poor recognition of different learning environments using traditional questionnaire-based methods that are time-consuming and do not respond dynamically to individual learner preferences and lack of absorption of learners' opinions and comments. The study's primary focus is to address learner performance in e-learning through intelligent techniques, which are essential for enhancing e-learning resources and learner experiences. The proposed method employed machine learning algorithms that are trained and tested using the Harvard person-course de-identified dataset. Machine Learning (ML) techniques such as Support vector machine (SVM), Decision tree (DT), Convolution neural network (CNN), Random Forest (RF), Naive Bayes (NB), Gradient boost (GB), Logistic regression (LR) and K- nearest neighbour (KNN) and ensemble model are used for predicting the percentage of the students. Further, based on the predicted outcome, the course is recommended to the students based on a hybrid collaborative filtering and content-based filtering model. The ensemble model yielded impressive results, with an accuracy of 0.993, recall of 1.00, precision of 0.991, and an F1-score of 1.00 in predicting the percentage of the students, and a proposed hybrid collaborative filtering and content-based filtering model achieved an accuracy of 0.99 in course recommendation. This study contributes to the increasing amount of knowledge about educational technology by providing useful information for instructors and developers to develop more successful and adaptive e-learning systems.