ABSTRACT
The school dropout rate in the IT field in Brazil was around 60.2%
in 2017, according to data released by the Ministry of Education’s
Higher Education Census. Considering this fact, this paper aims
to establish patterns regarding students who have abandoned the
Computer Engineering course at a federal institution in Brazil, using
a predictive model. An intelligent system capable of predicting
if a student will drop out or not based on his academic history
was proposed and implemented. The methodology for the system
development included four distinct phases. In the first phase, a data
extraction algorithm from academic histories was developed. The
second phase encompassed correlation analyses and the generation
of a decision tree to identify dropout indicators, using data from 180
students. Despite data limitations, the model achieved 78% accuracy.
In the third and fourth phase an Application Programming Interface
(API) and a system dashboard were implemented to achieve the
proposed goal.