Using Machine Learning to Estimate the School Effect on Student Trajectory Protection to Support Data-Driven Decision-Making

Turning points in students' educational trajectories are crucial moments that can significantly impact their academic and life paths. Some of these turning points are chronic absenteeism, grade retention, and school dropout, which significantly affect students' academic and socio-emotional outcomes, e.g., in educational achievement, stigma, decreased self-esteem, increased likelihood of suicide, antisocial behaviors, and detrimental effects on future economic stability and health. Therefore, there is a need for mechanisms to protect students' trajectories at the school level.

However, one of the underlying difficulties is assessing how well individual schools protect students’ trajectories to determine the support they need. To this end, we propose a machine-learning model that predicts trajectory interruption for a given year based on students' characteristics.

The model is calibrated so that the estimated probability corresponds to the actual incidence of the phenomenon. The calibration allows us to interpret the results at the school level in terms of the proportion of students who interrupt their trajectories each year to compare them with the actual results. We correct this difference by the general error of the model and express it in terms of the percentage of the predicted students suffering from trajectory disruption, determining which schools perform better (or worse) than expected, creating a School Trajectories Protection Index (STPI).

Using the SPTI, we identify differences at different levels (e.g., gender, nationality, grade, administrative dependence, and socioeconomic level) and their evolution on time that we make available in an interactive visualization tool that allows us to identify territories where educational trajectories are less protected, and need support and (or) intervention by authorities. Also, identifying schools with better protection than expected (especially on differences before and after the COVID-19 pandemic) is a case study to learn from their practices through qualitative studies and encourage collaboration network development for school improvement.