APLICANDO REDES NEURIAS ARTIFICIAIS PARA ESTUDO DE EVASÃO DE ENSINO SUPERIOR

Abstract

Dropping out from higher education is a critical issue that generates socioeconomic losses for society. Thus, this study aimed to identify the factors influencing dropout rates. Dropout profiles were investigated using Artificial Neural Networks (ANN) with students from the industrial engineering program at the Nova Andradina/MS campus. The research involved administering questionnaires to both students who remained in the program and those who dropped out. After processing the data, it was fed into a Multilayer Perceptron (MLP) neural network. To find the best model, the network was run 20 times, and performance indicators were analyzed. The results revealed that the primary decision-making factor for students is the Choice of Course, with 100% importance. When combined with other factors, the most significant weight is attributed to the choice of course. Secondarily, living arrangements in the city, including students who had to relocate to attend college, accounted for 78.6%. Other critical factors, with importance levels exceeding 40%, included satisfaction with the chosen course, the location of the college in relation to the student's residence in the city, and a lack of foundation in high school. The findings suggest multidimensional factors influencing dropout decisions and can serve as a tool for managers to make more informed decisions about critical factors, thus reducing dropout rates. Although the model was specifically applied to the industrial engineering program at the Nova Andradina campus, the methodology used in this study can be applied to other undergraduate programs.