<![CDATA[This systematic literature review critically examines the optimization of K-Means and Agglomerative Clustering algorithms for segmenting student academic performance. The study aims to map current research trends, evaluate the comparative application of Elbow and Silhouette validation methods, and identify significant gaps limiting the pedagogical utility of clustering outputs. The review employed the PRISMA protocol and conducted a rigorous search of the Scopus database, yielding 26 studies for the final qualitative synthesis and thematic analysis. Findings reveal a field dominated by quantitative, engineering-driven refinements and a clear trend towards algorithmic hybridization, particularly the use of metaheuristics to strengthen K-Means. While the Elbow and Silhouette methods are canonical, their application is often procedural rather than critically comparative. A core limitation is the pronounced theoretical deficit: clusters derived predominantly from secondary data remain statistically valid but pedagogically inert due to minimal integration with educational or learning sciences frameworks. The discussion underscores that the field's primary bottleneck is not algorithmic but interpretative, stemming from a methodological monoculture and a disconnect between computational output and actionable educational insight. The conclusion emphasizes the imperative for future research to develop explanatory, theory-guided models, employ mixed-methods and longitudinal designs, and address contextual Equity in data provenance to transform segmentation from a descriptive technique into a tool for genuinely understanding and supporting diverse learners]]>
