Predicting Student Academic Success Using Machine Learning Models: A Learning Analytics Approach in Higher Education

Abstract

Rapid deployment of digital learning technologies in the higher education sector has created immense amounts of educational data that could be leveraged to enhance student success and institutional effectiveness. Nevertheless, student dropout, poor academic performance, and lack of retention continue to plague universities across the world. In most cases, identification of academically struggling students is often late since existing models are largely reactive. Therefore, there is need for development of advanced learning analytics models that are able to forecast student performance in higher education institutions. The current study seeks to create an artificial neural network (ANN)-based learning analytics framework to predict student success in higher education institutions. A predictive analytical approach based on quantitatively evaluating a sample of 1,000 undergraduate students was used in the current study. Various attributes used to evaluate the students included demographic information, academic performance, LMS activity, and learning behaviors. Learning analytics indicators used in the model included previous GPA, attendance rate, assignment completion rate, quiz scores, logins per week, learning hours per week, discussion engagement, engagement index, interaction scores, and learning consistency. In the analysis, the model was validated and tested against accuracy, precision, recall, F1-score, ROC-AUC, confusion matrix, and cross validation tests. Results showed that accuracy, precision, recall, F1-Score, and ROC-AUC of the ANN model were 92.8%, 91.4%, 93.7%, 92.5%, and 0.96, respectively. Based on these outcomes, previous GPA, attendance rate, assignment completion rate, and various engagement indicators were found to be the strongest predictors of student success in college. On the theoretical front, contributions of this study include AI-assisted student performance and behavior prediction. Practically, a sophisticated warning system was developed in this study to assist in effective academic advisement and planning for student retention and academic improvement strategies.