Uma abordagem para predição de estudantes em risco utilizando algoritmos genéticos e mineração de dados: um estudo de caso com dados de um curso técnico a distância

Resumo

This paper demonstrates a non-traditional approach attempting to predict student dropout. For this, evolutionary system technics are used, which in this implementation seeks the optimization through the competition of six classifiers with initially random generated parameters among the possible ones for each classifier, the fitness function ranks the population at the end of each epoch. At last, the fittest individuals of each classifier are selected, and they compete with each other. The six classifiers, using the standard configuration, are compared and then contrasted with those obtained by the proposal. In this way, it was possible to obtain an improvement in the performance, with average gains ranging from 4% to 6%, in some cases up to 10%, depending on the metric.