Um modelo de dados para apoiar a mineração de dados educacionais na investigação de evasão de estudantes

• Autor(a) principal: MAGALHÃES JÚNIOR, Péricles Nogueira
• Data de Publicação: 2013
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UNIFACS
• Texto Completo: http://teste.tede.unifacs.br:8080/tede/handle/tede/571
• topic: Mineração de Dados Educacionais; Evasão de Estudantes; Modelo de Dados; Ciências Exatas e da Terra

The insertion of Distance Education in the Brazilian Law of Guidelines and Bases of Education (Law 9,394 of December 20, 1996), lifted the modality as one of the main strategies for inclusive education in the country, causing a growth of courses offer, with a diverse range of methodologies and educational formats. Extensive use of software solutions in the operation and management of these courses, such as learning management systems, produces a large volume of data related to the behavior of their students, unexplored raw material in decision making processes of these institutions. The data mining techniques and algorithms, for their ability to process large amounts of data to identify common patterns, have been used by researchers and managers of online courses supporting their decision processes. These papers, however, have little or no relation with each other regarding to their approaches and terminologies. The contribution of this dissertation is based on the construction and proposal of a data model that can join applicable indicators in various educational situations and can be used as a reference in future studies, providing greater uniformity of terminology and thus allowing comparative analyzes in different works. For the selection of entities and attributes that made up the proposed model, we conducted a literature review regarding the research conducted in the Educational Data Mining area and about the major conceptual models of student’s behavior analysis, focusing on the attrition phenomenon. Thus, we propose a reference model for Educational Data Mining applications and, focusing the attrition phenomenon, an evaluation study was conducted, showing that the model is applicable, allowing the identification of evidence of evasion of students, as well as to reduce the effort required for attributes selection and subsequent preparation of the data for data mining. The use of a reference model for future researches will enable the convergence of terms and concepts, providing greater exchange of experiences between researchers.