Abstract—This paper addresses the clustering challenges within big data for educational recommendation systems. Traditional Minimum Sum-of-Squares Clustering (MSSC) algorithms face critical limitations when applied to this domain, creating three distinct research gaps that this study aims to resolve: a persistent efficiency gap that hinders real-time application, a crucial interpretability gap that impedes pedagogical trust and adoption, and a significant stability gap caused by the dynamic and noisy nature of learner data. To bridge these gaps, we propose an improved MSSC framework integrating a distributed computing architecture, an Locality-Sensitive Hashing (LSH)-based dimensionality reduction stage, a hybrid Fuzzy C-Means (FCM)-based initialization strategy, and an adaptive feature weighting mechanism. Experimental results demonstrate significant, quantifiable improvements. The optimized algorithm reduces runtime by 37.2% and memory usage by 42.3%. It achieves a clustering accuracy of 92.3% and improves the Silhouette Coefficient by 15.8%. Furthermore, the algorithm demonstrates strong robustness against data loss and noise. These findings confirm that our proposed algorithm provides a more efficient, interpretable, and stable solution, offering reliable technical support for precise, personalized learning recommendations.
 
Keywords—educational recommendation system, minimum sum of squares clustering, big data processing, clustering algorithm, personalized learning


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).