Prediction of Students’ Academic Progression using Machine Learning

This paper is concerned with the development of an innovative method for predicting students’ academic performance and progression in their first semester of studies at higher education institutions in Australia. This is an essential factor contributing to the improvement of student retention which is considered as one of the most challenging problems within higher education sector, especially for private institutions. An artificial neural network was developed and trained by a training set to learn patterns of a group of 200 commencing students at Sydney Institute of Higher Education (SIHE). The neural network is subsequently tested with a testing set which consists of data from another group of 100 commencing students to examine the ability of predicting students’ achievement of learning outcome in their first semester at SIHE. With the best classification results of 85% Sensitivity and 69% Specificity on the training set, and 82% Sensitivity and 66% Specificity on the testing set, it is demonstrated that the developed neural network can successfully recognize new patterns of the new testing group of students and effectively identify students who are at risk of making unsatisfactory performance.