Predicting University Student Graduation Using Academic Performance and Machine Learning: A Systematic Literature Review

Predicting university student graduation is a beneficial tool for both students and institutions. With the help of this predictive capacity, students may make well-informed decisions about their academic and career paths, and institutions can proactively identify students who may not graduate and offer tailored support to ensure their success. The use of machine learning for predicting university student graduation has drawn more attention in recent years. Large datasets of student academic performance data can be used to train machine learning algorithms to identify patterns that are applicable in predicting future outcomes. In accordance with some studies, this approach predicts student graduation with an accuracy rate as high as 90%. Many systematic literature reviews (SLRs) have been conducted in this field, but there are still limitations, including not discussing the predictive models and algorithms used, a lack of coverage of the machine learning algorithms applied, small database coverage, keyword selection that does not cover all synonyms relevant to the investigation, and less specific data collection transparency. By delving into the limitations of existing SLRs on this topic, this research not only enhances the understanding of machine learning applications in forecasting student graduation but also fills a crucial gap in the literature. The inclusion of weaknesses in current SLRs provides a foundation for justifying the need for this study, emphasizing the necessity of a more nuanced and comprehensive review to advance the field and guide future research efforts in smart learning environments. This research conducts a thorough systematic review of the existing literature on machine learning-based student graduation prediction models from 70 journal articles from 2018 through 2023 that are pertinent. This review includes the various machine learning algorithms that have been implemented, the various academic performance data that was obtained from students, and the effectiveness of the models that have been developed. It also discusses the difficulties and potential advantages of utilizing machine learning to predict student graduation. The review indicates that the most common approach employed is the prediction of students' academic performance, which relies on data obtained from the Learning Management System and Student Information System. The primary data utilized for prediction purposes consists Student retention and time of academic and behavioral information. Among the various algorithms employed, Support Vector Machine and Random Forest are the most commonly utilized. This study makes a significant contribution to the advancement of learner modules within the smart learning environment.