A competitive learning approach to enhancing teacher effectiveness and student outcomes

Abstract

Machine learning has found extensive application and improvement in the field of education. Nevertheless, there remains a lack of research studies focusing on unsupervised learning within this domain. To address this gap, our study aims to investigate the relationship between teacher attributes and student achievement in Morocco while identifying regions requiring attention and intervention, using a novel clustering approach based on unsupervised competitive learning, specifically the 'Centroid neural network', to cluster Moroccan teachers based on their qualities and qualifications. Teacher qualities and qualifications are operationalized as initial teaching qualifications, completion of training programs, and employment status. To achieve our objective, we utilize the program for international student assessment (PISA) dataset, which provides comprehensive responses from individual students, including information on parental backgrounds, socio-economic positions, and school conditions. Additionally, we incorporate data from the teacher questionnaire, which encompasses background information, initial education, professional development, teaching practice, and teacher beliefs and attitudes. Consistent with previous research, our findings suggest that teachers' qualities and qualifications significantly influence student performance. Furthermore, our clustering approach identifies regions where there is a pronounced prevalence of attributes negatively impacting student achievement. Urging academicians to incorporate resilience-building measures into the design of policies in these regions to improve students' educational outcomes.