Comparação de diferentes configurações de bases de dados para a identificação precoce do risco de reprovação: o caso de uma disciplina semipresencial de Algoritmos e Programação

Resumo

This paper presents a comparison of the use of different datasets to generate models for predicting at-risk students. Four different datasets were tested together with three distinct classifiers in the context of an Introductory Programming Course. Initial results indicate that the use of a dataset composed by the counting of students interactions inside the Learning Management System during the period fo the course was the one that allowed the models to achieve the best performances.