Development of Machine Learning Models for Educational Performance Analysis and Student Success Prediction in Higher Education

Abstract With the quick evolution of digital technologies and data-driven educations, higher education institutions are accumulating large amounts of data related to students, such as academic achievements, attendance, evaluation results, demographical data, learning behaviours, and engagement levels. A good analysis of such information can enable educational institutions to uncover the factors that affect learner achievement and identify potential underachieving and withdrawal students. This research aims at building machine learning models for analysis of educational performance and prediction of student success in Higher education. The purpose of the study is to gather, pre-process and analyze data pertaining to student academic and behavioral data to uncover important patterns and factors that relate to student academic outcomes. A variety of machine learning methods will be explored to create predictive models to estimate academic performance and students at various academic risk levels. To ensure the reliability and prediction power of the developed models, suitable data preprocessing, feature selection, model training, model validation and model performance evaluation techniques will be used. Appropriate metrics for model performance will be chosen to accommodate the types of prediction tasks, including accuracy, precision, recall, F1-score, Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and the Area under the ROC Curve (AUC). The research will also involve the analysis of the most influential factors affecting students' achievement and offer some meaningful interpretation for academic decision making. The proposed approach can help HEIs to establish a system of early-warning to identify academically at-risk students and provide timely intervention to address the issue, including academic counselling, mentoring, remedial teaching and personalized learning support. In addition, the study aims to evaluate the performance of various machine learning models and identify the most appropriate models for this context of educational performance prediction in Higher Education. The intended result of this research is a machine learning framework that is reliable and interpretable with the ability to analyse educational performance and predict student success. The results could be used to inform evidence-based academic planning; personalized student support; increased retention; and better educational outcomes in higher education institutions.