Analysis of the Impact of the Pandemic on Social Inequalities in Enem 2019 and 2020 using Machine Learning

Abstract

ENEM measures the ability and knowledge of students who are in high school or have already completed it. With the scores obtained in the exam, the student can enroll in SISU, which is one way to enter public universities. During pandemic, the planning of schools, mainly public, was affected so that many students gave up taking the ENEM in 2020. To identify the profile of those enrolled in ENEM and verify which portion was most affected, this research analyze their social inequalities using data from ENEM 2019 and 2020 and machine learning methods. The methodology is based on cluster analysis where K-Means was applied and on performance classification where Random Forest, K-Nearest Neighbors, and MultiLayer Perceptron were used, and Select K-Best was used to select features. The results of the grouping generated two groups, one composed of subscribers with lower financial conditions and another with greater ones. In the classification, the MultiLayer Perceptron obtained an accuracy of 85.18% for 2019 and 83.63% for 2020. The results showed that the proposed methodology was able to identify the differences between the subscribers and classify their performance.