Using Virtual Learning Environment Data for the Development of Institutional Educational Policies

This paper describes the application of Data Science and Educational Mining techniques to data from 4529 students, seeking identify behavior patterns generate early predictive models at Universidad de la República del Uruguay. The use different sources (a Virtual Learning Environment, survey, academic system) discover most impactful variables linked student success. combination demonstrated a high power, achieving prediction rates with outstanding discrimination fourth week course. analysis showed that students more interactions inside Environment tended have success in their disciplines. results also revealed some relevant attributes influenced students’ success, such as number subjects was enrolled in, mother’s education, neighborhood. From emerged institutional policies, allocation computational resources for infrastructure its widespread use, development tools following trajectory detection at-risk failure. construction an interdisciplinary exchange bridge between sociology, science is significant contribution community may help constructing university educational policies.