Um modelo preditivo para diagnóstico de evasão baseado nas interações de alunos em fóruns de discussão

Resumo

This paper presents the results of a survey that is to create a predictive model of data mining in a Virtual Learning Environment (VLE). The obtained model aims to make the diagnosis of students evasion, based on their interactions in discussion forums, to help stakeholders in decision making. For this, experiments were performed with different sets of data being applied five classification algorithms and compared the performance of each. The results showed that the techniques based on the decision tree had the best performances and in some cases, achieved accuracy rate above 69%.