Uma abordagem computacional de predição de desempenho acadêmico de estudantes em cursos on-line de programação.

Abstract

There is a high failure rate in some initial disciplines in computing undergraduate courses, both in face-to-face and distance modalities, especially in programming and mathematics disciplines. Recently, it has been perceived as a tendency that the execution of such disciplines in the face-to-face modality has been associated to an online complementation, thus having, as it happens in distance education, a virtual learning environment linked, which enables interactions between students and teachers, potentially producing a large amount of data. In this context, a broad and important research problem arises that is how to apply data mining, via predictive modeling, to generate high quality information: correct, timely and useful, allowing, for example, effectively subsidize the process of pedagogical decisions to be taken by teachers, aiming to contribute to the reduction of the mentioned failure rate. In this sense, the overall objective of this research is to design and develop a predictive approach to identify, as soon as possible, students who may be at risk of failing or dropping introductory programming courses, taking into consideration to generate reliable and comprehensible prediction models. In this approach, we gave priority to white-box predictive models, appreciating that this type of model is potentially more adequate to offer comprehensibility in terms of information in the model. To evaluate the proposed approach, we conducted several empirical studies using academic data of students from a public university, as well as socioeconomic and demographic data. Thus, the obtained results showed the feasibility and effectiveness of the proposed approach, both with respect to the quality of the prediction, as well as in the indication of the influence of the selected attributes. Therefore, it was concluded that the model is interpretable and able to perform prediction with satisfactory accuracy within the comparative standards of the literature and, thus, it is useful to aid in pedagogical decision making by the teachers.