In the context of the educational quality evaluation measured through standardized tests, this article aims to select the context variables that have a greater contribution in the differentiation of the categories of the 2015 SIMCE math score, for eighth grade students of the region of La Araucanía, Chile. Based on a cross-sectional research, a supervised classification design was implemented, defining as an indicator of the SIMCE score the categories: Adequate, Elemental and Insufficient. Support Vector Machines (SVM) were trained to classify the students into these categories. Each student is represented by the variables obtained through the context questionnaires applied to parents and teachers. A Genetic Algorithm (GA) was applied to select which of these variables are the most relevant to discriminate the categories. The obtained results evidenced that SVM and GA are tools that can be applied in the educational field with good results to select variables. The most relevant variables coincide with those analyzed in the specialized literature: Student self-efficacy, Educational expectations of the parents, Educational level of the mother, family income, Index Student values, school social climate, classroom climate, among others.