Endogenous and exogenous factors for the modeling and prediction of the academic performance of university students

Abstract

The objective of this research is to design and implement an artificial neural network (ANN) model that allows predicting the academic performance of students at the Fabiola Salazar Leguía de Bagua National Intercultural University (UNIFSLB) in the Mathematics subject. This research presents a quantitative, non-experimental, projective and predictive approach; A dichotomous response questionnaire was developed to collect information on the factors that influence Academic Performance (AP). For the validation of the questionnaire, the expert judgment criteria was used, and for reliability the Kuder-Richarson test (reliability coefficient of 0.82). The study population was made up of 397 UNIFSLB students. The ANN model was designed in the MATLAB software, the model adjustment was performed taking into account the mean square error (0.27) and the weighted correlation coefficient during training, validation and testing (0.92%). The RNA model with the best prediction results is made up of three hidden layers (35-42-31 neurons in each hidden layer) and an output layer (1 neuron). It was concluded that it is possible to implement an RNA model with endogenous and exogenous factors to predict the AR of students.