Table 2.

Summary of the results of the classification AUCs for semisupervised learning methods under progressively increasing label rates. The learning performance of the graph convolutional network on the large data sets—that is, data sets other than Second Affiliated Hospital of Zhejiang University Colorectal Cancer—is unavailable due to memory limits.

Label Rate		5%, AUCa	10%, AUC	15%, AUC	20%, AUC	25%, AUC
University of California Irvine Machine Learning Repository Type 2 Diabetes 30-Day Readmission
	GSSLb	0.450	0.472	0.523	0.542	0.602
	LPc	0.475	0.475	0.564	0.585	0.566
	Proposed	0.929	0.979	0.964	0.930	0.924
Surveillance, Epidemiology, and End Results Ovarian Cancer
	GSSL	0.454	0.512	0.537	0.591	0.591
	LP	0.344	0.364	0.462	0.478	0.491
	Proposed	0.640	0.719	0.677	0.678	0.650
Surveillance, Epidemiology, and End Results Colorectal Cancer
	GSSL	0.525	0.527	0.447	0.585	0.578
	LP	0.540	0.532	0.512	0.540	0.513
	Proposed	0.595	0.652	0.640	0.581	0.590
Second Affiliated Hospital of Zhejiang University Colorectal Cancer
	GSSL	0.547	0.573	0.564	0.553	0.580
	LP	0.454	0.448	0.512	0.460	0.507
	GCNd	0.505	0.575	0.562	0.585	0.606
	Proposed	0.587	0.650	0.634	0.568	0.508

^aAUC: area under the receiver operating characteristics curve.

^bGSSL: graph-base semisupervised learning.

^cLP: label propagation.

^dGCN: graph convolutional network.