Aprendizado de máquina automático aplicado à predição da evasão no ensino superior

Academic dropout is a problem that affects many public and private university students in Brazil and around the world. Machine learning techniques have been used to mitigate the problem, but still require a lot of manual adjustments. We present in this work, a proposal of an automatic machine learni...

ver descrição completa

Detalhes bibliográficos
Autor: Barros, Bruno de Mattos
Tipo de documento: dissertação
Estado:Versão publicada
Data de publicação:2022
País:Brasil
Recursos:Universidade Federal de Goiás (UFG)
Repositório:Repositório Institucional da UFG
Idioma:português
OAI Identifier:oai:repositorio.bc.ufg.br:tede/12454
Acesso em linha:http://repositorio.bc.ufg.br/tede/handle/tede/12454
Access Level:Acceso aberto
Palavra-chave:Aprendizado de máquina automatizado
Segmentação temporal de dados
Predição da evasão acadêmica
Mineração de dados educacionais
Automated machine learning
Temporal data splitting
Academic dropout prediction
Educational data mining
CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
Descrição
Resumo:Academic dropout is a problem that affects many public and private university students in Brazil and around the world. Machine learning techniques have been used to mitigate the problem, but still require a lot of manual adjustments. We present in this work, a proposal of an automatic machine learning framework to predict academic dropout, with the goal of obtaining good results without the need for human intervention. This data processing framework includes the following stages: pre-processing, feature vector creation, data splitting into testing and training sets, clustering of data from different degrees for training, model selection, model parameter tunning and explainability. Additionally, we formalize temporal data splitting approaches for train and test datasets, as this task is not adequately addressed in most of the previous works.