<![CDATA[Purpose – This study develops an exploratory institution-level approach for prioritizing digital-learning improvements when only aggregate survey distributions are available.Methodology – A new secondary computational analysis used 35 five-point Likert indicators from a doctoral survey of 167 postgraduate respondents in selected Malaysian private-university contexts. Each indicator was represented by its five-category frequency distribution, transformed using square-root/Hellinger coordinates, and clustered with Fuzzy C-Means (FCM). Candidate solutions were compared using internal validity indices; the selected four-profile solution was benchmarked against favorable-direction means and favorable-response percentages; and 1,000 multinomial perturbations assessed aggregate-count stability.Findings – The four-cluster solution had MPC = .500, normalized PE = .487, Xie–Beni = .315, and silhouette = .345. Internet disruption was the clearest diagnostic concern, with 53.9% agreeing or strongly agreeing that access problems hindered online-learning completion. FCM ordering aligned with conventional ranking (Spearman ρ = .923 for mean rank; ρ = .888 for favorable-response rank). Perturbations produced mean ARI = .322, mean NMI = .475, mean retention = 62.8% (median 59.3%), and IA4 retention = 50.6%.Research limitations – The analysis clusters 35 indicators, not students. Aggregate data cannot recover respondent-level relationships, and perturbation tests aggregate-count robustness rather than respondent-level or population stability. IA4 is treated as a diagnostic outlier signal, not a robust singleton class.Originality – The study provides an exploratory, auditable soft-prioritization approach that preserves five-category response profiles and graded membership while avoiding unsupported claims of superiority over simpler ranking.]]>
Copyrights © 2026
