Investigating student engagement, learning, and attendance during the COVID-19 disruption: a machine learning analysis of online, hybrid, and offline learning modes

Abstract The COVID-19 pandemic profoundly impacted educational systems worldwide, catalysing a transformational shift towards digital teaching and learning at all levels. In response, schools and institutions supported teachers to enhance their online teaching, focusing on technological and pedagogical aspects, with many providing the necessary resources to facilitate online teaching and learning during the disruption. This study aims to investigate key factors that affected student learning, engagement, and attendance during the COVID-19 disruption. This causal-comparative study aims to identify the variables that best predict student outcomes across different teaching modes. Using data from the Responses to Educational Disruption Survey, which included 21,063 students, 15,004 teachers, and 1,581 principals from 11 countries, the analysis employs the k-nearest neighbours machine learning algorithm to identify influential factors. Students receiving education through different modes are commonly affected by their access to learning resources, feedback received, and whether their teachers meet and collaborate with others teaching the same grade level. For online and hybrid modes, reducing curricula to its core elements became more prominent, while for the offline mode, teachers’ professional training, learning materials creation, and peer collaboration were more significant.