Table 3.

Graphics processing unit memory consumption of the proposed method against that of typical semisupervised learning algorithms on 4 data sets. The graph convolutional network is not suitable for data sets other than Second Affiliated Hospital of Zhejiang University; therefore, only one result is shown.

	UCI-T2Da, MB	SEER-OVCb, MB	SEER-CRCc, MB	SAHZU-CRCd, MB
Graph-based semisupervised learning (CPUe)	1374	770	1260	297
Label propagation (CPU)	1200	702	1263	257
Graph convolutional network	Out of memory	Out of memory	Out of memory	732
Proposed	336	345	330	332

^aUCI-T2D: University of California Irvine Machine Learning Repository Type 2 Diabetes 30-Day Readmission.

^bSEER-OVC: Surveillance, Epidemiology, and End Results–Ovarian Cancer.

^cSEER-CRC: Surveillance, Epidemiology, and End Results–Colorectal Cancer.

^dSAHZU-CRC: Second Affiliated Hospital of Zhejiang University.

^eCPU: central processing unit.