Table 8. Results of model validation on link prediction task for MAG-CS dataset (accuracy metric lies between (0,1) and higher value means better results).

Bold corresponds to the best metric for each dataset.

	GBM	LR	RF
GRAREP (Cao, Lu & Xu, 2015)	0.5986 ± 0.0047	0.5626 ± 0.0016	0.5998 ± 0.0025
HOPE (Ou et al., 2016)	0.566 ± 0.0017	0.5275 ± 0.0025	0.6007 ± 0.0027
NODE2VEC (Grover & Leskovec, 2016)	0.578 ± 0.0023	0.5425 ± 0.0015	0.6137 ± 0.0031
WALKLETS (Perozzi, Kulkarni & Skiena, 2016)	0.5798 ± 0.0024	0.5647 ± 0.0015	0.6077 ± 0.0027
GCN (Kipf & Welling, 2016a)	0.8486 ± 0.0022	0.6553 ± 0.0014	0.8772 ± 0.0012
GAT (Veličković et al., 2017)	0.7293 ± 0.0041	0.5632 ± 0.0024	0.7524 ± 0.0027
GCN (NF)	0.7253 ± 0.0012	0.697 ± 0.0014	0.7261 ± 0.0017
GAT (NF)	0.5015 ± 0.0009	0.5015 ± 0.0009	0.5015 ± 0.0009