Online learning and teaching increased in 2020, driven by the COVID-19 pandemic. As many researchers attempted to understand the impact stress had on the emotional behaviours and academic performance of students, most studies explored these pre- and during-COVID behaviours in the context of brick and mortar institutions transitioning to online delivery. There is an opportunity to compare the experiences of students in the MOOC environment in this period, particularly in terms of the difference of engagement, semantics and sentiment/stress behaviours in 2019 and 2020. In this study, we use a dataset from AdelaideX between this time period to identify the most significant features that impact student outcomes. Where previous machine learning approaches used singular features such as student interaction or sentiment in discussion forum posts, we incorporate three feature categories of engagement, semantics and sentiment/stress in an ensemble model is based on voting and stacked methods to determining the relationship between them and academic performance. From our results, we discover that sentiment/stress played little part in academic performance and was relatively unchanged in online courses in this dataset between 2019 and 2020. We present two individual student cases to further contextualise our findings.
Bibliographical noteFunding Information:
The authors would like to thank Ali Ogilvie and the entire Online Programs Team from The University of Adelaide for approval of data usage and ongoing support of this project.
The authors wish to acknowledge the Human Research Ethics Committee approval from The University of Southern Queensland, reference number H20REA137.The authors would like to thank Ali Ogilvie and the entire Online Programs Team from The University of Adelaide for approval of data usage and ongoing support of this project.
© 2023 The Author(s)
- Performance modelling
- Sentiment analysis