Authors:
Maicon Dall’Agnol
1
;
Leandro Rondado de Souza
1
;
Renan de Padua
2
;
Veronica Oliveira de Carvalho
1
and
Solange Oliveira Rezende
2
Affiliations:
1
Universidade Estadual Paulista (Unesp), Instituto de Geociências e Ciências Exatas, Rio Claro, Brazil
;
2
Universidade de São Paulo (USP), Instituto de Ciências Matemáticas e de Computação, São Carlos, Brazil
Keyword(s):
Dropout, Association Rules, Network, C4.5.
Abstract:
Dropout is a critical problem that has been studied by data mining methods. The most widely used algorithm
in this context is C4.5. However, the understanding of the reasons why a student dropout is a result of its
representation. As C4.5 is a greedy algorithm, it is difficult to visualize, for example, items that are dominants
and determinants with respect to a specific class. An alternative is to use association rules (ARs), since
they exploit the search space more broadly. However, in the dropout context, few works use them. (Padua
et al., 2018) proposed an approach, named ExARN, that structures, prunes and analyzes a set of ARs to build
candidate hypotheses. Considering the above, the goal of this work is to treat the dropout problem through
ExARN as it provides a complementary view to what is commonly used in the literature, i.e., classification
through C4.5. As contributions we have: (a) complementary views are important and, therefore, should be
used more often when the focus
is to understand the domain, not only classify; (b) the use of ARs through
ExARN may reveal interesting correlations that may help to understand the problem of dropping out.
(More)