Combinando Técnicas de Mineração de Dados para Melhorar a Detecção de Indicadores de Evasão Universitária

Resumo

The student' dropout is one of the major problems faced by public universities in Brazil since generates financial, social and academic loss. In this work, we present a methodology which combines different data mining techniques that aims: (1) creating predictive models to identify students who are at risk of dropout; and (2) establishing the relevance of the attributes related to the phenomenon and thus contribute to its prevention. We evaluate the methodology applying it in real information regarding students of the Federal University of São João Del Rei, performing both a global analysis of the data and a fragmented analysis for each Knowledge Area. The predictive models generated show alarming numbers for possible evasions in 2019, especially in Engineering; Agrarian Sciences; Exact and Earth Sciences; and Applied Social Sciences. Regarding the identified attributes, the academic indicators were the most relevant to understanding the students dropout, a factor that places the institutional manager as the main figure to combat this phenomenon.