Recomendações de recursos educacionais baseadas em aprendizagem de máquina para autorregulação da aprendizagem

Abstract

The growth of distance education in Brazil is stimulating the development of solutions to motivate students, minimize evasion and enhance student performance. Data gathered in virtual learning environments (AVA) have been considered to extract behavioral patterns and use them to improve performance. AVAs have been improved to process information and suggest strategies that enhance the level of student learning. This work develops a recommendation model based on auto-regulation patterns of learning in AVAs with the employment of machine learning and data mining algorithms. A software platform to recommend education activities was built to (1) analyze the performance of students according to a score; (2) extract behavior characteristics which influence performance, and (3) recommend actions to improve student performance. Experiments conducted with a database of more than 30.000 students of a Brazilian university, with several performance metrics, showed the proposed solution was capable of capturing learning profiles with over 0.95 AUROC (Area under the Roc Curve).