Proposed hybrid Model in Online Education

Abstract

The advancement of technology powering e-learning has brought numerous benefits, including consistency, scalability, cost reduction, and improved usability. However, there are also challenges that need to be addressed. Here are some key considerations for enhancing the technology powering e-learning. Artificial intelligence has revolutionized the field of e-learning and created tremendous opportunities for education Storage, servers, software systems, databases, online management systems, and apps are examples of such resources. This paper aims to forecast students' adaptability to online education using predictive machine learning (ML) models, including Logistic Regression, Decision tree, Random Forest, AdaBoost, ANN. The dataset utilized for this study was sourced from Kaggle and comprised 1205 high school to college students. The research encompasses several stages of data analysis, including data preprocessing, model training, testing, and validation. Multiple performance metrics such as accuracy, specificity, sensitivity, F1 score, and precision were employed to assess the effectiveness of each model. The findings demonstrate that all five models exhibit considerable predictive capabilities. Notably, decision tree and hybrid models outperformed the others, achieving an impressive accuracy rate of 92%. Consequently, it is recommended to utilize these two models, RF and XGB, for predicting students' adaptability levels in online education due to their superior predictive accuracy. Additionally, the Logistic regression, KNN, and AdaBoost, ANN models also yielded respectable performance levels, achieving accuracy rates of 77.48%, 83.77 ,74.17% and 91.06% respectively. In summary, this study underscores the superiority of RF and XGB models in delivering higher prediction accuracy, aligning with similar research endeavours employing ML techniques to forecast adaptability levels.