Predicting teachers’ research reading: A machine learning approach

Abstract

In addition to pre- and in-service teacher education programmes, teachers’ autonomous reading of content related to their work contributes significantly to their professional development. This study investigated the factors that influenced the professional reading of 10,469 language teachers in the 2018 dataset of the Programme for International Student Assessment (PISA). Two machine learning models – logistic regression and Support Vector Machines (SVM) – were used to classify light and heavy readers. Nineteen variables related to teachers, including various aspects of their life, education and instructional practices, were used as predictors for classification. The results indicate that the two models had very similar accuracy scores around 65%. Moreover, the length of the reading texts that teachers assign to their students, instruction of reading comprehension strategies, and teachers’ own general reading habits were found to be the most important predictors of professional reading time.