Utilização da mineração de dados para identificar a evasão nos cursos EaD do Instituto Federal de Educação, Ciência e Tecnologia Farroupilha

• Autor(a) principal: Sonnenstrahl, Thiago Siqueira
• Data de Publicação: 2020
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Manancial - Repositório Digital da UFSM
• Texto Completo: http://repositorio.ufsm.br/handle/1/23307
• topic: Educação a distância; Evasão; Mineração de dados educacionais; Distance education; Dropout; Educational data mining; CNPQ::CIENCIAS HUMANAS::EDUCACAO

The Farroupilha Federal Institute is a component of the Federal Network of Basic, Professional, Technical, and Technological Education and strives for the presence and success of its students in accordance with Institutional Development Plan (IDP) 2019/2026. Managing the performance of students in a virtual teaching and learning environment (VLE) is of fundamental importance to reduce dropout and failure rates in distance education (DE) courses. Thus, by using Educational Data Mining (EDM) and assessing student interaction on the VTLE, this study aimed to analyze possible dropouts in DE courses at the Farroupilha Federal Institute by providing strategic data for educational managers of the institution. The development of the present study was divided into four distinct stages and based on a bibliographic review employing a qualitative and quantitative approach. The first stage sought, through exploratory research, dropout data and other information from the distance education department of the Farroupilha Federal Institute. The second stage took place with a bibliographic review on dropout rates in distance education. The third step was data mining and the evaluation of results. The fourth and last stage consisted of a qualitative analysis of mining data as a way of guiding the institution to make decisions within the scope of the Distance Education Department while considering student interactions on the VTLE. The study was developed by performing three experiments using interactions on the VLE Moodle of two classes of a subsequent distance education course. Each experiment consisted of a class and the third experiment was the unification of the data in a single set. As a result, the mining of experiment 3, which joined the data of both classes and was obtained with the Random Forest algorithm, showed that the score rate was higher than 88%. The best attributes that performed the prediction were task visualization and material visualization. The master's dissertation presented here is in the line of research of the Development of Educational Technology in Networks, part of the Graduate Program in Educational Technology in Networks and generated as products the text presented here and the created EDM strategy.