Cargando...
Exploring patterns in intercultural bilingual education in Peru using machine learning
Author(s)
Jenny Tarrillo Vásquez
Lina Mamani
Hasnain Iftikhar
Ronny Ivan Gonzales Medina
Date Issued
5 de junio de 2026
Type
Article
Volume
11
Abstract
Introduction Intercultural Bilingual Education (IBE) in Peru aims to provide culturally and linguistically relevant instruction for Indigenous students; however, its implementation continues to face structural, linguistic, and resource-related challenges. This study identified latent patterns in IBE implementation using machine learning techniques and compared the predictive performance of different classification algorithms. Methods A dataset comprising 84,558 institutional records and 17 variables was analyzed. After data preprocessing, including missing-value imputation, categorical encoding, and dimensionality reduction, multi-class classification models were developed to predict five IBE implementation scenarios. Eight machine learning algorithms were evaluated under different training-test configurations (80/20, 50/50, 25/75, and 10/90). Model performance was assessed using accuracy, precision, recall, weighted F1-score, and five-fold cross-validation. Results Ensemble-based methods consistently outperformed linear and distance-based classifiers. Gradient Boosting achieved the best performance (accuracy ≈ 0.77; weighted F1-score = 0.74), followed by Random Forest and K-Nearest Neighbors. Cross-validation confirmed model robustness, with mean accuracies ranging from 0.766 to 0.774 and low standard deviations (< 0.007). Discussion The findings demonstrate that machine learning models can effectively capture complex and nonlinear relationships in educational systems characterized by linguistic and institutional diversity. The proposed framework offers a scalable and reproducible approach for evidence-based policy development, supporting data-driven strategies to improve equity and quality in intercultural bilingual education.
Keywords