Modelo predictivo de exito académico aplicando algoritmos de aprendizaje de máquina sobre interacciones en el TEC Digital

Abstract

Machine learning and data mining have many applications in the educational field. In particular, in the design of predictive models from massive data records related to learning process. In the present work, several predictive models of academic success are designed and evaluated, in order to create support tools for early intervention on cases of academic failure. These Ad hoc models were designed for the TEC Digital’s educational platform. The machine learning algorithms used are Logistic Regression, Support Vector Machines and Neural Networks. The input dataset to the compared models consists of student’s interactions within TEC Digital’s educational platform. The best results obtained correspond to the algorithm of Neural Networks, but it weren’t found the parameters nor the levels of complexity to adjust models that predict with an accuracy greater than 80 %, the students who are most likely to fail a course.