Answers are grounded in this paper and cite their sources.
Selecting a cloud service from many competing providers is a multi-criteria decision problem in which the Quality of Service (QoS) criteria are often strongly correlated. Methods such as AHP, BWM and TOPSIS treat these criteria as independent and need many pairwise comparisons. An earlier study by the authors combined Principal Component Analysis (PCA) with a two-level Best-Only Method (BOM), in which an expert also compared every provider under every criterion, at a cost of 95 judgements for eight criteria and twelve providers. The present paper keeps the PCA stage, the BOM weighting of the criteria and the variance-weighted utility of that study, and removes the provider-level elicitation. BOM weights are applied directly to the standardised QoS data, so only n − 1 judgements are needed (7 instead of 95) and the effort no longer grows with the number of providers. The two versions are compared on the same twelve QWS services and under the same two preference profiles. Under the reliability-focused profile of the earlier study, both select the same three leading services, and their rankings are significantly correlated (Spearman's ρ = 0.80); under the response-time profile, the agreement is weaker (ρ = 0.56). In all four common-basis sensitivity settings, the single-level version is the more stable (mean ρ 0.93–0.98 against 0.82–0.94). Its ranking also agrees with PCA-BWM (ρ = 0.98) and Weighted PCA (ρ = 0.94). The final component scores are shown to be uncorrelated only for equal weights.
No takes yet. Share an insight, caveat, or question.
Patel et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: