Soluções tecnológicas para o problema da evasão universitária, sob a óptica de ferramentas de inteligência artificial

Abstract

This research expands the debate on the issue of school dropout in Brazilian Higher Education. The losses generated by the interruption in the cycle of studies are numerous and their consequence is in human development, causing more and more socioeconomic and cultural incompatibilities in the country. In this context, it is essential to develop innovative methods, processes and solutions, capable of helping people. Educational Institutions in the fight against dropout. Therefore, the present work aims, in general, to know the context of school dropout in Brazilian higher education; map and discuss technological advances in the field of Intellectual Property and Industrial Property; to present the application of artificial intelligence techniques in an academic context. And finally, the development of an innovative and intelligent solution, capable of predicting students with a tendency to drop out and / or graduate. As investigative material, the academic data from the face-to-face courses at the State University of Tocantins (Unitins) were explored and research bases were used, such as theses, dissertations and scientific articles. The study demonstrated that school dropout is a complex phenomenon, since this problem is not restricted to Brazilian Educational Institutions, a phenomenon that also afflicts the international scenario. It is noted that several studies are being developed by researchers, increasingly contributing with solutions and means to overcome the problem. The research also mapped on the technological bases both in Brazil and in the world the quantitative of technological solutions referring to patents and registrations of computer programs. By the numbers, we see a very low quantity, with few records and, in addition, many of these technologies have been canceled or extinguished. When it comes to the use of Academic Data Mining, for the discovery of knowledge about school dropout data, the research used Unitins' academic data for the 12 on-site courses. The areas used for artificial intelligence were Educational Data Mining (MDE), Machine Learning (AM) using the random forest algorithm. The methodological proposal allowed the proactive and accurate identification of students at risk of dropping out or with a predisposition to graduate. The Intelligent System aims to provide information about school dropout easily, from the first periods of the course. With the consistency of the results, it allowed the University of the State of Tocantins (Unitins) to obtain 95.21% accuracy between the periods, thus certifying the reliability, precision, robustness of the random forest algorithm. Providing management with proactive and assertive decision making, anticipating the reversal of the evasion situation.