Approaches to Predicting Educational Problems: A Systematic Mapping

  • Paulo Mello Silva UFPE
  • Fernando da Fonseca UFPE
  • Roberta Fagundes UPE


Currently, educational data present alarming situations regarding the problems faced by educational institutions, such as low levels of learning, reading and writing performance, high levels of failure and dropout among others. These problems represent a major obstacle for institutions seeking to provide society with quality education. Given this context, it is essential to identify the factors associated with these problems. To minimize the occurrence of educational problems, several types of research use approaches/techniques, such as Educational Data Mining (EDM), Learning Analytics (LA), Machine Learning (LM). These approaches/techniques can analyze educational data generated in teaching-learning environments by applying data mining tasks such as prediction (regression and classification) grouping or associating data, to make intrinsic knowledge discoveries in Dice. The main objective of this work is to identify through a Systematic Mapping of Literature, the main approaches / predictive techniques, which have been used to predict educational problems in teaching-learning environments. Also, identify the main factors that affect the teaching-learning process of students. The results of this study indicate that educational performance has been affected by factors related to the following aspects : academic, demographic, socioeconomic, technological and behavioral. The most commonly used techniques for predicting educational problems in the literature are classification, regression, and clustering.
Palavras-chave: Educational Problems, Educational Prediction and Educational Data Mining
Como Citar

Selecione um Formato
SILVA, Paulo Mello; DA FONSECA, Fernando; FAGUNDES, Roberta. Approaches to Predicting Educational Problems: A Systematic Mapping. In: SIMPÓSIO BRASILEIRO DE SISTEMAS DE INFORMAÇÃO (SBSI), 16. , 2020, Evento Online. Anais [...]. Porto Alegre: Sociedade Brasileira de Computação, 2020 . DOI: