Skip to main content
. 2022 Apr 5;13:810362. doi: 10.3389/fpsyt.2022.810362

FIGURE 1.

FIGURE 1

Receiver Operator Characteristics (ROC) curves for classification of schizophrenia patients and controls based on different combinations of features using logistics (A), random forest (B), and XGBoost algorithm (C). NSF subset, Neurocognitive Selected Features subset include IMM, LAN, ATT, DEM, INT-C, INT-W features; ESF subset, Electrophysiological Selected Features subset include PSC-PPI, PSS-PPI, Abs-T, Abs-A, Abs-AFp/AO, Abs-(D+T)/(A+B), Rel-D, Rel-T, Rel-A/B, DFA-A, DFA-B. ASF set, All Selected Features set include NSF subset and ESF subset. The red, green and blue lines show the ROC curves for the NSF subset, ESF subset and ASF set, respectively. The AUC of logistics models based on NSF subset, ESF subset, ASF set was 89.88%, 90.84%, 92.54%. The AUC of random forest models based on NSF subset, ESF subset, ASF set was 96.59 91.88%, 97.36%. The AUC of XGBoost models based on NSF subset, ESF subset, ASF set was 93.99%, 90.52%, 97.91%.