Análise do perfil de estudantes de graduação baseada nas suas experiências e percepções sobre a universidade: uma abordagem por técnicas de clusterização

Abstract

With the challenge of understanding student profiles within the university context, data mining has been increasingly used as a means to generate useful information for decision makers. In this sense, the present work sought to apply clustering techniques on data obtained through a questionnaire applied to students at a public university. To this end, a questionnaire was prepared with 39 quantitative and qualitative questions, containing questions about the student and their perceptions regarding the university, in which they assess satisfaction and loyalty, according to the CSAT (Customer Satisfaction) and NPS (Net Promoter Score), respectively. Through it, three clustering techniques were applied, namely: K-Means; Fuzzy C-Means; and Hierarchical Method. The data set was divided in two, one for questions in which the student evaluates the university experience and another for a personal question. The results were evaluated by the silhouette method, which returned the best values for K-Means in both cases, with scores of 0.181 (with two clusters) and 0.498 (with three clusters), respectively. However, for the case of three clusters, Fuzzy C-Means managed to achieve a better balance of instances between the groups, bringing the opportunity to analyze in parallel with K-Means. Thus, the differences between the groups were analyzed, resulting in at least five distinct profiles. Two of them differed, mainly, for being “promoters” or “neutral/detractors” in relation to the university, and they were characterized by facing greater responsibilities and a more stressful life situation. The other profiles differed, in particular, by the relationship they have with the family, by attending a psychologist or psychiatrist, and by the amount of situations of stress, anxiety or lack of motivation they face during graduation.