Estudo e desenvolvimento de um detector automático de comportamento Gaming the System em resolução de problemas de programação

Abstract

This work is related to a study on a type of behavior, considered undesirable, identified in some students, when faced with a situation of solving programming problems. Thus, we invested in the development of a solution to automatically detect the presence of the mentioned behavior. Thus, firstly, an appropriate environment for data collection was developed, followed by the preparation of these data to be used by supervised machine learning algorithms, considering a classificationtask,allowingthedetectionofbehavior. The development of the detector involved exploring, evaluating and comparing different classification algorithms. As one of the results, we found that the MLP algorithm performed better, with an accuracy of 0.84.