Student performance predictive models using LMS data in Primary Schools

Abstract

Plan Ceibal is a public policy implemented in Uruguay, it is part of the global initiative One Laptop per Child (OLPC, 2005). The basic feature is providing every student and teacher in primary school with a laptop or tablet and internet access. Different data sets were combined, students and teachers activities registered in the Learning Management System (LMS) and student's performance in national standardized tests. Data were used to compute student's engagement indexes, combining motivation, creativity, velocity and performance. Statistical models were used to determine key drivers of LMS use, this is relevant to define educational policies based on evidence. Models for LMS use are fitted for several regional levels. Additionally, statistical learning methods were fitted to predict student's performance in national standardized test using as predictor variables different constructed usage indexes from the LMS platform. A major challenge was how to deal with sub-grouping data structure into machine learning algorithms, usually developed for independent observations. Initial results suggest school district is the main driver of the technology usage in the classroom.