Looking through the fog of remote Zoom teaching: a case study of at-risk student prediction

Abstract

Identification of students who are at-risk of failing or dropping out from a course is a key part of instructional remediation for student retention. The data-driven machine learning approach has proven to be effective in utilising student information to make the prediction. The Zoom video conferencing platform, which has become widely adopted to replace in-person teaching and learning in the COVID-19 pandemic, poses a challenge to building effective at-risk student prediction model. Extracting information about students is made difficult by increased capacity to control self-disclosure and the manipulation of online communication. The case study described in the paper aims to find out the feasibility of at-risk student prediction in Zoom teaching and the capacity of engineering informative features based on the polling function. A number of prediction scenarios were defined and the performance of the corresponding models and the effectiveness of various machine learning algorithm were evaluated. It was found that formative assessment features were useful for prediction scenarios earlier in the course, and summative assessment features gave accurate predictions towards the end. The findings have filled the knowledge gap of at-risk student prediction in Zoom teaching.