This study aimed to employ machine learning techniques to uncover the pivotal determinants influencing the reading proficiency of fourth-grade students across 65 regions, as participants in the PIRLS 2021 assessment. The primary objective was to discern and assess key factors at the student, family and school levels that predict high and low reading performance among these students. Utilising a machine learning approach, this research analysed data from 204,176 fourth-grade students encompassing 122 independent variables. The Support Vector Machine (SVM) algorithm was employed to effectively differentiate between students with high and low reading performance based on the identification of 16 crucial contextual factors. The results revealed that the most influential factors predominantly resided at the family level, encompassing socioeconomic variables. These factors pertained to the provision of personalised study environments, facilitated through access to an internet connection, individual study desks, dedicated study rooms, an assortment of books and personal smartphones. At the student level, significant factors included reading motivation, gender and age. Meanwhile, school-level determinants encompassed aspects such as ineffective classroom regulations, absenteeism rates, the presence of libraries and the availability of digital learning resources.