Resumen

CARDOZO, Santiago; SILVEIRA, Adrián  y  FONSECA, Bruno. Early Detection of School Risk. Prediction of Trajectories of School Backwardness in Primary Education in Uruguay Using Machine Learning Techniques. Rev. latinoam. estud. educ. [online]. 2022, vol.52, n.2, pp.297-326.  Epub 09-Feb-2022. ISSN 2448-878X.  https://doi.org/10.48102/rlee.2022.52.2.391.

The work uses computational learning techniques (machine learning) to estimate school risk during the first three years of the trajectory in primary education of a cohort of Uruguayan students. We use three analysis techniques (logistic regression, Bayesian networks, and classification trees) to identify the risk of school trajectories based on the repetition of at least one year, based on a set of factors before the transition to the first year of primary school. These factors range from the students' socio-sanitary conditions at their birth to their family and educational situation at the end of their pre-primary schooling. In particular, the analysis focuses on the predictive power of the skills captured by the Early Childhood Assessment (EIT) that is applied close to the end of initial education, around five years of age. The results suggest that the skills captured by EIT manage to identify in advance the majority of children with educational risk. The levels of precision and sensitivity of the models that include this factor show the potential of early warning systems to detect and prevent situations of “school failure”.

Palabras llave : child development; school risk; machine learning; repetition.

        · resumen en Español     · texto en Español     · Español ( pdf )