Sistema para la detección temprana de anomalías en la evaluación usando técnicas de aprendizaje automático

Zusammenfassung

One of the most important processes in almost all uni- versity education models is evaluation. The criteria es- tablished in a subject guide how the student’s final grade is obtained. Therefore, it is important to con- tinuously monitor the student’s learning process and grades, thus allowing the detection of anomalies to pro- ceed with an immediate intervention to correct the si- tuation. Typically, the first university courses have a high number of students, which is detrimental to the tracking that can be done by the teacher. In this paper, we propose an approach to predict the next grade of a student in a certain activity, so that the teacher is no- tified in case the actual grade is different enough from the predicted one. To this end, an experimental study of 24 artificial intelligence algorithms, selecting the most suitable ones for our case of study. The experimental results show the goodness of the proposed approach, and that the algorithms based on support vector machi- nes or those of extreme gradient augmentation are the ones that best fit the considered data.