Identification of factors that affect the academic performance of high school students in Peru through a machine learning algorithm

Abstract

The Peruvian Ministry of Education annually conducts the Student Census Evaluation (ECE, for its acronym in Spanish) to evaluate the level of learning achievement in the subjects of mathematics, reading and science and technology, both in public and private schools. The results are classified as Before beginning, Beginning, In process or Satisfactory. According to the results of the ECE 2019, it is observed that the academic performance achieved in the area of mathematics presents the highest percentage of students at the Satisfactory level (17.7%); however, in turn, said field of study is also the one that groups the highest percentage of students at the Before beginning level (33.0%).

Considering the aforementioned, this research aims to identify those variables that affect the learning achievements in mathematics of high school students. Thus, for the proposed analysis, a classification model was built for each of the mentioned levels, through an ensemble machine learning algorithm that uses the gradient boosting method. As a result of the modeling, the importance of the variables analyzed was obtained, which finally identified those that have greater relevance in the prediction of the classification of each level of learning achievement.