Enhancing the Early Prediction of Learners Performance in a Virtual Learning Environment

Educational institutions have widely adopted virtual learning environments (VLEs) in contemporary education. The limits of students’ location are no longer a problem because they can learn from anywhere and anytime based on this approach. Hence, by forecasting students’ performance in VLEs, educational institutions can enhance their online offerings and provide quality online learning content. This is not possible without taking into account various features that may have a great influence on students’ academic accomplishments. The present paper intends to predict students’ performance in an online platform. Four classifiers are utilized in the proposed model. This research integrates the whole dataset for science and social science modules. Moreover, to improve the model's prediction accuracy several steps are followed. First, new features are generated based on the available features namely, the total number of clicks before, the total number of clicks after, engagement, and average. Second, the model’s hyperparameters are adjusted using the random search optimizer, whereas a feature selection approach is performed to choose the maximum influential features. The experimental results showed that the prediction accuracy is significantly enhanced based on the procedure proposed in this research. The suggested model successfully provides an early prediction of students’ performance with an average accuracy of 84%. The outcomes of this research are discussed further to highlight its possible implications on theory and practice.