A Data Mining Approach Applied to the High School National Examination: Analysis of Aspects of Candidates to Brazilian Universities

In many college courses in several countries are used exams in a national scale, such as Gaokao, in China, Scholastic Aptitude Test - SAT and the American College Testing - ACT in the United States of American, Yuksekogretime Gecis Sinavi – YGS in Turkey, among others. This paper examines microdata from the High School National Examination (ENEM) database from Brazil. The database has 8,627,367 records, 166 attributes, and all experiments were performed based on the Spark architecture. The objective of this work is to examine microdata of the ENEM database applying data mining algorithms and creating an approach to handle big data and to predict the profile of those enrolled in ENEM. Through the standards found by the data mining algorithms with classification algorithms, it was possible to observe that family income, access to information, profession, and academic history of the parents were directly related to the performance of the candidates. And with a rules induction algorithm, it was possible to identify the patterns presented in each of the regions of Brazil, such as common characteristics when a candidate was approved and when not, essential factors as disciplines and particular characteristics of each region. This approach also enables the execution of large volumes of data in a simplified computational structure.