Graph convolutional and attention models for entity classification in multilayer networks

Applied Network Science

Table 2 Summary of structural characteristics of the network datasets: type of the network, number of layers (l), number of entities, or actors (a), number of nodes (n), number of edges (e), density mean/SD over the layers, node degree mean/SD over the layers, average entity frequency (aef), homophily, and number of classes (c)

Network	Type	l	a	n	e	density	degree	aef (%)	homophily	c
Balance	Social	4	625	2500	155,000	\(0.199 \pm 0.0\)	\(124.0 \pm 0.0\)	100	\(0.507 \pm 0.0\)	3
CKM-Social	Social	3	241	723	2740	\(0.016 \pm 0.001\)	\(7.580 \pm 0.632\)	100	\(1.0 \pm 0.0\)	4
Congress	Vote	16	435	6960	707,872	\(0.496 \pm 0.025\)	\(203.411 \pm 10.86\)	100	\(0.670 \pm 0.115\)	2
DKPol	Political	3	490	839	20,226	\(0.07 \pm 0.008\)	\(28.991 \pm 44.115\)	57.07	\(0.518 \pm 0.282\)	10
Leskovec-Ng	Collaboration, Temporal	4	191	214	536	\(0.166 \pm 0.151\)	\(4.088 \pm 1.454\)	28.01	\(0.994 \pm 0.012\)	2
Starwars	Interaction	6	92	157	494	\(0.253 \pm 0.045\)	\(6.117\pm 0.835\)	28.44	\(0.571 \pm 0.067\)	3
Terrorist-Noordin	Terrorist	4	79	227	1822	\(0.129 \pm 0.05\)	\(13.829 \pm 9.087\)	71.83	\(0.780 \pm 0.118\)	2
Terrorist-status	Terrorist	4	79	227	1822	\(0.129 \pm 0.05\)	\(13.829 \pm 9.087\)	71.83	\(0.469 \pm 0.069\)	3
Vickers	Social	3	29	87	740	\(0.304 \pm 0.122\)	\(17.011 \pm 6.854\)	100	\(0.763 \pm 0.088\)	2
Koumbia-2-mpx	Geospatial	2	2246	2246	6011	\(0.002 \pm 0.001\)	\(5.355 \pm 0.017\)	50	\(0.814 \pm 0.004\)	2
Koumbia-5-mpx	Geospatial	5	2246	3776	12,870	\(0.005 \pm 0.002\)	\(6.751 \pm 0.448\)	33.62	\(0.827 \pm 0.097\)	2
Koumbia-10-mpx	Geospatial	10	2246	7191	27,433	\(0.006 \pm 0.003\)	\(7.567 \pm 0.392\)	32.02	\(0.826 \pm 0.091\)	2
Koumbia-15-mpx	Geospatial	15	2246	10,591	43,285	\(0.007 \pm 0.003\)	\(8.103 \pm 0.458\)	31.43	\(0.826 \pm 0.090\)	2
Koumbia-20-mpx	Geospatial	20	2246	14,069	59,968	\(0.008 \pm 0.07\)	\(8.447 \pm 0.489\)	31.32	\(0.823 \pm 0.089\)	2