Anticipating Student Abandonment and Failure: Predictive Models in High School Settings

Addressing the issue of high school non-completion poses a crucial challenge for contemporary education. This research introduces a machine learning-based methodology to identify students at risk of failure and abandonment in a specific Brazilian state, aiming to establish an early warning system utilizing academic, socioeconomic, and performance indicators for proactive interventions. The methodology followed here ensures the explainability of predictions and guards against bias in relation to certain features. The analysis of data from 79,165 students resulted in the creation of six accurate classification models, with accuracy rates ranging from 69.4% to 92.7%. This underscores the methodology’s effectiveness in identifying at-risk students, highlighting its potential to alleviate failure and abandonment. The implementation of this methodology could positively influence proactive educational policies and enhance educational metrics within the state.