Implementación de técnicas de machine learning y creación de una red neuronal artificial para la predicción del rendimiento académico de estudiantes en ambientes universitarios que usan e-learning y streaming

Abstract

This work describes the implementation of machine learning

(ML) techniques: Random Forest, Xtreme Boosting Gradient,

Support Vector Machine, K-Nearest-Neighbor and Logistic

Regression as well as the creation of an artificial neural

network (ANN), which were compared to determine the

technique that can learn to predict with greater accuracy, the

low academic performance of university students, to improve

the mechanisms of e-learning and streaming that help them

raise academic performance. The e-learning methodology was

established for the first time in the late 1990s, however, since

the Covid-19 pandemic, it has established itself as the best

alternative to traditional education, placing it as a benchmark

worldwide. One of the concerns in the university environment

where this study was carried out is to be able to determine

the impact that virtual teaching has had compared to face

to-face teaching, since there are factors (gender, number

of children, sex, age, type of study) that could influence the

academic performance of students. Using the classification

metrics within the comparative process, it was determined that

among the implemented ML techniques, the XGBoost reached

78.4% accuracy, but was surpassed by the artificial neural

network (ANN) that learned to predict with 82.4%. of accuracy.

Due to the above, the use of the artificial neural network is

recommended for the prediction of the academic performance

of university students since, in addition, with its massive

predictions can be made due to its high processing capacity.