Abstract:

This article presents a framework that allows moving from the simple grading of students in virtual courses to a real evaluation process in a massive and autonomous way. For this purpose, variables related to a holistic evaluation are defined such as live class visualization, pre-recorded class visualization, performance in the class test, performance in the deliverables to create the portfolio. We have collected the data of approximately 300 students participating during 4 months of a virtual course of Digital Art, to subsequently train several machine learning and deep learning models that allow predicting the performance of 100 students of the same course, identifying possible cases of risk, favoring the decision making regarding the accompaniment of students and the development of the course. The models made with machine learning showed a high level of reliability when predicting the position of each student at the end of the course; no substantial improvement was found when using deep learning.

Keywords:

Scoring, virtual course assessment, deep learning, machine learning.