Predicting Student Success for Programming Courses in a Fully Online Learning Environment

Abstract

The emergence of online learning environments is important for teaching programming courses. In this study, demographic and performance-related data from two programming courses of a fully online learning platform, UniSA Online, were explored. Statistically significant features were identified using Varian Inflation Factor and Chi-Square test. Four prediction models were trained and tested using four sets of features: demographic, performance, statistically significant features, and all available features. The model trained using demographic features yielded an accuracy of 45.45%. The models trained usind performance-related features, statistically significant features, and all features yielded an accuracy of 86.86%, 86.53%, and 86.53%, respectively. This highlights the importance of performance-related data in predicting student success outcomes in learning programming via a fully online learning environment.