Hierarchical Interpretability in ML: A Novel Approach to Explaining Student Adaptability Predictions Using H-LIME

In this study, we introduce Hierarchical Local Interpretable Model-agnostic Explanations (H LIME), an innovative extension of the LIME technique, designed to enhance the interpretability of machine learning models by offering explanations at multiple levels of data hierarchy. This study focuses on predicting student adaptability across various educational contexts, a critical factor in personalized education and student support strategies. H-LIME aggregates local explanations to deliver insights across different data hierarchy levels, including institution type, location, and edu cational level. Applied to a dataset with diverse student adaptability features, H-LIME revealed key insights, such as the consistent influence of education level and class duration across hierar chical contexts and the varying impact of factors such as network type and financial condition depending on specific subgroups. These findings underscore the utility of H-LIME in uncovering complex patterns in educational data, providing a valuable tool for stakeholders to implement targeted interventions based on hierarchical insights. This study establishes the effectiveness of H-LIME in the educational domain and highlights its potential application in other fields where data hierarchies are critical. This approach offers a scalable solution for enhancing model interpretability, thereby facilitating more informed decision-making in data-driven environments.