Abstract:

This paper presents the application of a predictive model created from a neural network, trained with data obtained from the behavior of students belonging to 2 different cohorts (n=300 students) of the technical training program in digital arts for film and video games during the first semester of the year 2022. The data were obtained with data mining techniques implemented in the self-created platform, specifically the performance in quizzes, participation in live and pre-recorded classes and interaction with peers in forums were analyzed. The final model was applied to 2 new cohorts (n=220). The first one was formed by a group selected through a careful selection process and the second one was formed by students selected only for their motivation to participate in the program, both groups were in the first month of study of the program and based on their results it was possible to carry out preventive and corrective actions on the program methodologies and the accompaniment of students identified by the model as possible dropouts. It was found that the group formed by students selected from various filters, such as interviews, technical and profiling tests, the model predicts a success rate of 95% while the group that was selected based solely on their motivation the model predicts a success rate of 43%.

Keywords:

Scoring, virtual course assessment, deep learning, artificial Intelligence, education.