Publicação

Using Machine Learning Models to predict high school student’s Academic Achievement

Ver documento

Detalhes bibliográficos
Resumo:Understanding student dropout has become increasingly relevant given the growing importance of educated people in today’s workforce. Therefore, predicting a student’s academic achievement (AA), whether he/she passes the academic year or not, can prove crucial to assisting teachers and competent decision-makersto create measures to help retain and eventually reduce academic abandonment. To address such issues, this paper utilizes Machine Learning (ML) models to obtain accurate predictions of essentially every student’s AA in Portuguese public high schools using data from the Portuguese Ministry of Education and understand what are the drivers of AA that most affect the predictive abilities of said models. Our results show that Random Forest and XGBoost have similar levels of accuracy, however, the latter displayed slightly better predictions. Regarding the most influential AA drivers, previous retention, gender, and the location of the student’s school were the ones that showed the greatest effect on the XGBoost model’s ability to accurately predict the student’s success. Several suggestions are made to educational stakeholders on the results of this study.
Autores principais:Quintino, Afonso João Mendes
Assunto:Education Academic Achievement Artificial Intelligence Machine Learning SDG 4 - Quality education SDG 5 - Gender equality SDG 10 - Reduced inequalities
Ano:2024
País:Portugal
Tipo de documento:dissertação de mestrado
Tipo de acesso:acesso aberto
Instituição associada:Universidade Nova de Lisboa
Idioma:inglês
Origem:Repositório Institucional da UNL
Descrição
Resumo:Understanding student dropout has become increasingly relevant given the growing importance of educated people in today’s workforce. Therefore, predicting a student’s academic achievement (AA), whether he/she passes the academic year or not, can prove crucial to assisting teachers and competent decision-makersto create measures to help retain and eventually reduce academic abandonment. To address such issues, this paper utilizes Machine Learning (ML) models to obtain accurate predictions of essentially every student’s AA in Portuguese public high schools using data from the Portuguese Ministry of Education and understand what are the drivers of AA that most affect the predictive abilities of said models. Our results show that Random Forest and XGBoost have similar levels of accuracy, however, the latter displayed slightly better predictions. Regarding the most influential AA drivers, previous retention, gender, and the location of the student’s school were the ones that showed the greatest effect on the XGBoost model’s ability to accurately predict the student’s success. Several suggestions are made to educational stakeholders on the results of this study.