Establishment and Analysis of a Combined Diagnostic Model of Polycystic Ovary Syndrome with Random Forest and Artificial Neural Network

Figure 6

Performance evaluation of different classification models by the area under the receiver operating characteristic (ROC) curves and their AUC values. (a) In microarray ComBat dataset2 (GSE43264 and GSE124226, ), neuralPCOS achieved superior performance (AUC: 0.7273), compared to the other two methods: EC-PCOS (AUC: 0.5985) and GC-PCOS (AUC: 0.5227). (b) In GSE84958 RNA-seq validation data (adipose, ), neuralPCOS achieved an AUC of 0.6488, EC-PCOS (AUC: 0.5770) and GC-PCOS (AUC: 0.7530). The optimal threshold values were labeled at inflection points, and the sensitivities and specificities were listed in the bracket.

(a)

(b)