Towards Predictive Modeling of High School Students’ Graduation Mentions Using Classification Techniques

Abstract This study develops a machine learning–based system to classify high school students’ graduation mentions using common-core grades. Using data from 72,020 students in the Guelmim-Oued Noun region to solve the problem of orientation students, the XGBoost algorithm was trained for multi-class classification across literary, scientific, and technical streams. The model achieved high accuracy: 92% for the literary stream, 89% for the scientific stream, and 90% for the technical stream. These results demonstrate the robustness of XGBoost for educational analytics. The proposed classifier can support early orientation decisions and help identify students requiring academic intervention. This work contributes to integrating data-driven tools into Morocco’s educational guidance system.