Modelo matemático para predecir el grado de deserción de los estudiantes en el Instituto Superior Tecnológico Bolívar

Abstract

Student desertion is an indicator of delay in education in any country in the world and this occurs with a greater impact in Ecuador. For this reason, it is important to take into account the research work that is presented through the following thesis. The purpose of this research is to predict the desertion of students from the Bolívar Higher Technological Institute in the city of Ambato. The Institute's archives show that 1,036 students were enrolled in the May-October 2022 semester in five majors.

The calculated sample size is 280 surveys using a confidence level of 95% and a margin of error of 5%. The surveys were applied to two groups, the first to: 115 students who for any reason abandoned their studies in any semester but returned to continue their studies. A second group of 165 students who did not drop out at any time, used Google forms for the survey through multiple choice questions . Due to having an asymmetric dataset, it is necessary to balance the data so that the models to be used work correctly.

In the present work, two algorithms are used: Support Vector Machines (SVM) and Feed-forward Artificial Neural Networks. When comparing the results obtained based on the F1 score, it was determined that the best algorithm was Feed-forward Artificial Neural Networks with an F-score value of 82.2% while in Support Vector Machines it yielded a value of 72.5 %.