Uso da técnica de mineração de dados como uma ferramenta de gestão da evasão no ensino superior

Abstract

For this study, the proposed approach is an ex post facto applied research, of the quantitative and descriptive type, based on the information about the students existing in the database of the Academic System (SISCAD) of UFTM. This research aims to identify a pattern of information about students that may be related to the possibility of dropout of higher education. With this study, it was possible to create an automated tool, based on the technique of data mining using artificial neural networks, capable of identifying students with a tendency to evasion. The model was able to identify 63.8% of the students who dropout. Among all the students analyzed during the validation of the model, it was able to correctly identify the situation of 70.5% of the students (between evaded and not evaded). It was also possible to define that the model can identify the students a tendency to evasion 36 days in advance of the occurrence of evasion, on average. An alert generation mechanism has been proposed, through the production of a report that can be made available directly in the academic management tool (SISCAD). In this way, the Institution’s agents, such as the course coordinators and the servers of the Pro-Rector’s Office for Community and Student Affairs (PROACE) of the UFTM, will be able to act together with the students with this tendency of evasion and try to reverse it.