Um modelo descritivo para auxiliar o acompanhamento da evasão escolar nos cursos técnicos e superiores no Instituto Federal do Rio Grande do Norte - Campus do São Gonçalo do Amarante

Abstract

The Federal Institutes of Education, Science and Technology (IFs) have as their goal the social transformation of the region in which they operate, contributing to the economic development of the country. Solutions that improve the performance or prevent a greater loss of resources of these institutions has become increasingly necessary. A proposal is to reduce dropout since the received budget is related to the amount of active enrollment they have. The São Gonçalo do Amarante (SGA) campus of the Federal Institute of Rio Grande do Norte (IFRN) presented in its active regular classroom courses from 2015 to 2017 a dropout average rates of 13.4% for the technical level courses and 31.5% for higher ones. These values are above the target values set in the Institution's Strategic Plan for Permanence and Success and the National Education Plan 2014-2024. Given this scenario, this research aims to propose a descriptive model to assist the management of the IFRN SGA campus in decisionmaking to reduce the dropout of students of the Integrated Computer Technician (TII) and undergraduate program of Computer Network Technology (TRC). For this purpose, data related to academic, socioeconomic, demographic, participation in research and extension projects and participation in student assistance programs were extracted from the institution's academic system. These data were organized into categories according to the factors that influence dropout. Data mining techniques following the CRISP-DM methodology was used to discover implicit patterns and possible correlations between them by using Orange software. Related work that focused on understanding or proposing solutions to the problem of dropout using data mining techniques were analyzed and some of them were able to predict dropout of students with rates higher than 85%. By using the Exploratory Data Analysis (EDA) from data visualization techniques, this study observed that students with a variety of characteristics may evade. However, the found patterns indicate that the probability of dropout in the TII course is higher, for example, when the student enters the institution over 16 years old or failing more than four times in subjects. To evaluate which attributes are more related to each other, decision tree algorithm was used because through it is possible to verify the rules that are being used to classify a student's situation as evaded or not. Based on this, the Lince tool was developed, which enables the student to be monitored with a focus on the rules and attributes that were detected as most relevant to dropout and, thus, assisting in proposing actions to combat it.