Beyond Administrative Compliance: Evidence from Accreditation Data in Primary Schools in the Archipelago Region

Educational quality is a multidimensional construct shaped by pedagogical, social, and institutional factors, yet the relative contribution of these dimensions to school accreditation outcomes remains insufficiently understood. This study aimed to examine the educational and institutional factors associated with accreditation outcomes in elementary schools in Maluku Province, Indonesia, using an interpretable machine learning approach. The study utilized accreditation data collected between 2020 and 2023, comprising twenty educational and institutional indicators representing student development, teaching and learning processes, school climate, leadership, and management. Multiple regression-based machine learning models were developed and compared, and the best-performing model was subsequently interpreted using feature importance and SHAP analyses. The results revealed that student competence, student character, learning quality, teacher professional development, and learning climate were the most influential factors associated with accreditation outcomes. In contrast, managerial dimensions such as financial management, facilities management, and community engagement showed relatively smaller contributions. These findings indicate that school quality is primarily shaped by social and educational processes occurring within schools rather than by administrative arrangements alone. The study highlights the value of interpretable machine learning for educational evaluation and provides evidence that may support data-informed policies and school improvement initiatives, particularly in geographically diverse educational contexts.