Using data mining to predict students' academic success

• Autor(a) principal: André Filipe Roque Silva
• Data de Publicação: 2016
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: https://hdl.handle.net/10216/110241
• topic: Engenharia electrotécnica, electrónica e informática; Electrical engineering, Electronic engineering, Information engineering

Currently, universities record large amounts of data about students. Despite its potential to inform decisions regarding the allocation of resources and efforts, this information tends to be overlooked. Educational data mining is a recent research field that focuses on the use of data mining techniques to transform large volumes of educational data into useful and relevant knowledge that can improve the educational processes and decisions. This work intends to propose a set of three models. The first two will use the information available at start of the first semester and second semester, respectively, of the first year of the student's academic path to predict the academic success of the students enrolled at FEUP at the end of that semester. The third model will use the information available at the end of the first year to predict the academic performance of the students enrolled at FEUP at the end of their degree. At the same time, this work also intends to identify the factors that are the most critical to these models. The results of this project could allow college principals to identify students in need of more pedagogical support, as well as students with a high probability of excelling in their studies. It could also allow them to focus their attention on the critical aspects, by implementing mechanisms that tackle students' difficulties. The first step of the developed work consists of data cleaning and preparation processes that normalize the data retrieved from the university's information system and the definition of the derived variables. The second step is an empirical analysis of different algorithms with interpretable models in order to identify the models best suited for the tasks of predicting academic success regarding their performance with the available entry data. Thirdly, an analysis of the generated models will be presented, along with the identification of their main predictive attributes.