Mineração de dados para identificação do perfil de evasão de alunos da UFC - Campus Quixadá

Abstract

The high dropout rates in undergraduate courses have become increasingly worrying in Brazil, this problem has generated losses both for the country and for students and universities. In this context, the objective of this study is to identify students with a tendency to dropout from the UFC - Campus Quixadá/CE, by means of data mining techniques and using historical data of students, in which experiments were conducted with two distinct scenarios, the first scenario having the total number of records with the division of records by unbalanced class and the second scenario containing a sample of records with the division between balanced classes. The results obtained show that the potential students to escape can be identified with hit rates of up to 99% in the first scenario and in the second scenario of up to 95.5%. By means of these results, it is intended to help the managers of the institution in the decision making and in the elaboration of policies to mitigate school dropout