Hierarchical Fuzzy Feature Clustering With Iterative Ensemble Learning for Student Performance Prediction.

Accurately predicting student performance while preserving interpretability remains a key challenge in educational data mining, largely due to the heterogeneous, high-dimensional, and interdependent nature of student learning-related data. Existing models typically treat these features independently, neglecting their fuzzy correlations and hierarchical structures, thereby limiting their prediction accuracy and transparency. To address these issues, we propose a hierarchical fuzzy feature clustering with iterative ensemble learning (HFFC-IEL) model. We first design a hierarchical fuzzy feature clustering (HFFC) algorithm that integrates agglomerative hierarchical clustering with fuzzy c-means (FCMs) refinement to automatically group correlated features into interpretable fuzzy feature clusters. For each cluster, an iterative ensemble learning (IEL) module is designed, where two complementary base estimators jointly exploit primary (within-cluster) and auxiliary (cross-cluster) feature subsets to progressively reduce residual errors. The cluster-level predictions are subsequently aggregated via an averaging ensemble to obtain the final output. Extensive experiments on seven case studies demonstrate that the proposed HFFC-IEL consistently outperforms the machine learning baselines and recent state-of-the-art models. Statistical analyses further confirm that integrating HFFC with iterative refinement significantly enhances predictive performance. These findings highlight the potential of HFFC-IEL for advancing accurate, interpretable, and personalized educational analytics.