Modeling the presence of higher education institutions in global rankings: evidence from Brazil using SciVal indicators

Abstract This study examines how SciVal bibliometric indicators discriminate between Brazilian higher education institutions (HEIs) that appear in major global university rankings and those that remain absent. Using data from 87 Brazilian HEIs and SciVal indicators covering the period 2018–2025, we apply an integrated analytical framework combining K-means clustering and supervised models, including logistic regression, decision trees (J48), and support vector regression (SMOreg), with class balancing and stratified tenfold cross-validation. The results indicate that a limited set of SciVal indicators — particularly the share of publications in high-impact journals, citation-based impact measures, and collaboration patterns — exhibits substantial discriminative power for ranking presence, achieving classification accuracy of approximately 80%. However, because bibliometric indicators are incorporated directly or indirectly into ranking methodologies, the findings should be interpreted as evidence of discriminative overlap rather than causal determinants of inclusion. Accordingly, model-derived decision thresholds are sample- and period-specific and should not be treated as stable benchmarks. The clustering analysis further identifies an intermediate or “emerging” institutional profile characterized by moderate bibliometric strength but persistent absence from the analyzed rankings, suggesting that bibliometric signals may be necessary but not sufficient for international ranking visibility. The implications are discussed considering contextual heterogeneity, institutional missions, and the potential unintended consequences of ranking-oriented strategies.