Um estudo exploratório sobre os aspectos que influenciam na eficiência da predição de sucesso acadêmico de alunos de programação introdutória

Abstract

There is a lack of intelligent tools to assist teachers in the analysis of the large amount of data generated on the students of a particular course, allowing to measure the level of learning of the students or to assist in the identification of those who are not achieving a good performance in learning . To mitigate this type of gap, there are currently several studies in the area of Data Mining and Machine Learning, addressing the process of obtaining knowledge to assist in decision making. Thus, this area has been exploited to support teaching and research, allowing the extraction of relevant information, among the available data, to support decision-making by educators, mainly as well as managers. This work seeks to explore the factors, within the prediction process, that influence the efficiency of the already known techniques of prediction of success within the educational scope. For this purpose, four different but correlated approaches were applied to a real database to measure the efficiency of each technique. These approaches take into account the particular characteristics of each student, as well as to explore the efficiency of each technique related to the training sample and test, specific characteristics of the decision tree algorithms and temporal aspects of the prediction.