Mineração de dados aplicada na identificação da propensão à evasão na universidade

Abstract

Dropout propensity is a problem difficult to identify in Universities, since there are several factors that may influence its occurrence, and students from different universities may have different reasons for dropping out. Identifying such cases one by one is an impractical task, since it requires dealing with large amounts of data. The main objective of this work is to identify features that may contribute students to drop out from the Federal University of Pampa (UNIPAMPA), using data collected from the university academic system. The present work was limited to the scope of students from the Computer Engineering undergraduate program of UNIPAMPA. The collected datasets were treated and analyzed in order to apply Knowledge Discovery in Databases (KDD) techniques. This task was done with the help of a tool called Weka, in which data from the university is mined. The intention here is to discover features associated with dropouts in order to identify situations of students in risk and then to find ways to create and optimize the prevention of these dropouts. Through the application of different classification algorithms, association rules and clustering techniques, it was possible to verify an association between the poor academic performance and the drop-outs.