Redes neuronales de apoyo en la docencia bajo Moodle para predicción de deserción y reprobación escolar

Artificial Intelligence (AI) designs computer tools that simulate human intelligence processes, including learning, reasoning, and selfcorrection. In the educational field, AI facilitates the evaluation of student progress thro ugh neural networks that adjust tests to previously achieved achievements. The purpose of this work was to design an Artificial Neural Network (ANN) that will analyze the performance of students in the subject of Biomedical Informatics I at the Facultad de Medicina of the UNAM in order to estimate the probabilities of passing, failing, and dropping out. Materials and methods: A neural network was designed in Python, consisting of 3 fully connected layers, using performance information from approximately 2000 students in the Moodle platform for Biomedical Informatics I during 2019, 2020, and 2021 at the Facultad de Medicina of UNAM. The neural network was trained to estimate the probable performance of the students, executing 2500 epochs with a mean absolute error level of 0.0158. A kappa accuracy rate of 0.947 was obtained through a confusión matrix when classifying between passing and failing grades. For visualization, a web system with a traffic light system was designed to allow evaluation and teaching coordinators from the Department of Biomedical Informatics to consult the results obtained from the network. Discussion: The network must continue to be fed with controlled data, meaning data that is specifically worked for the network’s training, since the quality of the data in the input directly determines the quality of the data obtained at the end of the process, affecting the network’s objective attainment. Conclusion: There is still much to learn and work to be done, but the interdisciplinary effort will allow us to improve our system and ultimately support the educational process of future doctors in the best way possible. Keywords: Neural networks; education; Moodle; dropout; failure; AI; machine learning.