Desenvolvimento de modelos para estimar principais características socioeconômicas para as notas de redação e matemática para ENEM 2018

Abstract

Predictive analytics is na advanced analytical technique that uses data, algorithms and Machine Learning to anticipate trends and make business projections. Using the data collected by the Exame Nacional do Ensino Médio (ENEM) in 2018, this work aims to analyze the socioeconomic data of the participants, in order to improve regression models to predict the Writing and Mathematics scores of the participants. Predictive models were developed using Machine Learning algorithms (Neural Network, Random Forest, Decision Tree, Linear Regression). After applying metrics to evaluate predictive models, it was found that the Neural Network model performed better in predicting math scores, while Linear Regression in predicting writing scores. Some socioeconomic features on type of school, sex, computer, father’s occupation, monthly family income, which are influential in the prediction of Writing and Mathematics scores, stood out.