Skip to main content
. 2020 Mar 27;14:191. doi: 10.3389/fnins.2020.00191

FIGURE 2.

FIGURE 2

Feature selection process of using the support vector machine–classifier-based recursive feature elimination (SVM-RFE) algorithm with 5,653 dynamic functional connectivity (DFC)-based variables and 248 static functional connectivity (SFC)-based variables, respectively. Panel (A) represents the curve of the area under the curve (AUC) values using the top n features from the DFC matrices, and the red dot in the local magnification of the curve stands for the highest AUC value of 0.9975 achieved by the top 28 features. Panel (B) displays the curve of the AUC values using the top n features from the SFC matrix, and the blue dot in the local magnification of the curve shows the highest AUC value of 0.8746 achieved by the top 14 features.