Fig. 5: Including hierarchical organisation features in a Support Vector Machine classification yields better accuracy than traditional functional or effective connectivity features.
From: Reconfiguration of functional brain hierarchy in schizophrenia
Comparison of the accuracy obtained using a SVM algorithm with different inputs (FC, EC, FDT and EC + FDT) to identify the most informative metrics. a Violin plots display the distribution of SVM accuracy across 1000 k-folds using FC, EC, FDT and EC + FDT data as inputs. b Confusion matrices illustrate the overall accuracy achieved corresponding to the distributions in (a), presented as percentages represented on a color scale. P stands for positive, N for negative.
