Previsão da evasão estudantil em disciplinas introdutórias de programação por meio de mineração de dados sociodemográficos

Abstract

Student dropout is characterized as an exclusion process from educational environment determined by motivational, structural, socioeconomic factors, internal and external to educational institutions. Dropout occurrences in introductory computer programming classes, known as CS1, is a challenge often observed in courses in sciences and engineering. The present paper aims to build a student‘s predicting dropout model in CS1 classes of these courses with the use of sociodemographic data and suitable to the application of this model even at the beginning of each academic period. The applied methodology was based on the data mining process CRISP-DM (Cross Industry Standard Process of Data Mining), with adaptations to the educational environment, to extract knowledge and build the dropout predictive model using a student sociodemographic data dimension. In order to validate the proposed methodology, experiments were carried out with data from former students of CS1 from the science and engineering courses at UFAM. The prediction of dropout students in these classes proved to be feasible, being built a predictive model easily adaptable using the AdaBoost classifier, allowing the engagement of more efficient institutional and pedagogical initiatives to combat evasion in an attempt that this probability does not materialize.