Abstract Global educational inequality is influenced by socio-economic development, especially in low-income and conflict-affected regions. Predicting such inequality allows researchers and policymakers to design better education policies and resource allocation strategies. The Belief Rule Base (BRB) is used as an interpretable model that incorporates expert knowledge, making it suitable for these predictions. However, BRBs face challenges, including rule explosion due to excessive feature information and a lack of standardized hierarchical modeling. In addition, parameters in optimization processes are often influenced by randomness, causing them to deviate from expert knowledge, which reduces interpretability. To address these issues, we adopt the interpretable hierarchical confidence rule base (HBRB-I) model, which enables self-organized construction of hierarchical structures through a multilayer tree structure (MTS), and a new optimization scheme to improve accuracy while retaining interpretability. Experimental results show that the HBRB-I model achieves high accuracy and robustness in predicting global educational inequality, which also provides data support for equitable and sustainable distribution of global educational resources. The interpretability of the model allows policymakers to develop forward-looking education policies that are aligned with sustainable development goals.