Fig. 2From: An early prediction model for gestational diabetes mellitus based on metabolomic biomarkersMetabolite selection through Random Forest analysis. A Predictive accuracy graph through Random Forest analysis. The best model that fits the highest predictive accuracy range is a model that combines 10 variables with a predictive accuracy of 85.4% (red circle). B The top 15 metabolites with the highest average importance to identify GDM cases from controlsBack to article page