Decision Trees Applied to Academic Performance in Mathematics in Basic Primary

In this paper, the results of applying the decision trees classification technique to discover patterns of academic performance in mathematics in the Saber 5th tests are presented. This test was presented by fifth grade students from Colombian elementary schools, in 2017. The CRISP-DM methodology was used. Socioeconomic, academic, and institutional information was selected from the ICFES databases. A minable dataset was obtained using data cleaning and transformation techniques. The performance of several decision tree algorithms was analyzed with the Weka tool. The AdaboostM1 algorithm with J48 was selected for obtaining the best results. The model was built with this algorithm. Among the most important factors in the patterns discovered, associated with academic performance in Mathematics are the nature and location of the school, whether or not the student failed a grade, her age group, and the mother's education. The knowledge discovered in this research, constitutes quality information for the decision-making of the MEN, the secretaries of education and the directors of the basic primary educational institutions in the definition of improvement plans that result in the quality of education of elementary school in Colombia.