Multi-class classification to track students' academic outcome

Despite various measured taken by the universities and colleges, the number of college and university dropouts is still on the rise. About 19.5% of students in the U.K drop out of their colleges every year. Thus, the identification of tentative dropouts and failure-prone students can act as an early warning system while also assisting teachers in analysing the need to streamline their course according to the weaker students' need. Learners' academic performance majorly depend upon their demographic factors as well as some other learning-based factors.

Objective: This paper focuses on classifying students as Pass, Fail or Dropout, based on their demographic and learning features. Learner's demographic factors like their like Age, Gender, Highest Qualification, Index of Multiple Deficiency, Physical Disability and learning factors like student's past assessment grades, their interaction with Virtual Learning Environment have been considered for this study.

Dataset: A publicly available anonymised student data from Open University named as Open University Learning Analytics Dataset OULAD has been used for this study.

Methodology: To carry out this multi-class classification, five different machine learning and deep learning algorithms, Artificial Neural Networks (ANN), Random Forest, Decision Trees, XGBoost, Support Vector Machine (SVM), have been implemented.

Results: Accuracy has been considered as a metric of evaluation for the models. ANN performed the best with 78.08% accuracy.