LPbyCD: a new scalable and interpretable approach for Link Prediction via Community Detection in bipartite networks

Applied Network Science

Table 7 The LPbyCD performance comparison with state of the art approaches on generic data sets

Data set	Approach	Performance				Time, h
Data set	Approach	AUC	\(\sigma\)	AUPR	\(\sigma\)	cpu	exact
MovieLens	LPbyCD(Spectral part.^a)	0.89	0.00	0.38	0.00	1759	233
	LPbyCD(Louvain^a)	0.89	0.00	0.34	0.01	32	32
	DTIP_MDHN	0.96	0.00	0.69	0.00	211	30
	DTiGEMS+	0.94	0.00	0.56	0.00	24	37
	ALADIN^b	0.93	0.00	0.51	0.00	1076^b	4672^b
	BLM-NII	0.89	0.01	0.29	0.02	609	20
	WNN-GIP^b	0.78	0.07	0.21	0.08	3667^b	152^b
	NetLapRLS\(^b\)	0.92	0.00	0.47	0.00	1267\(^b\)	17\(^b\)
Unicodelang	LPbyCD(Spectral part.^a)	0.79	0.02	0.17	0.02	91	4
	LPbyCD(Louvain^a)	0.80	0.01	0.17	0.03	1	1
	DTIP_MDHN	0.97	0.01	0.71	0.03	5	1
	DTiGEMS+	0.79	0.02	0.23	0.02	2	2
	ALADIN^c	0.85^c	0.01	0.20^c	0.03	25	541
	BLM-NII^c	0.83^c	0.01	0.07^c	0.00	229	6
	WNN-GIP	0.70	0.01	0.06	0.01	411	12
	NetLapRLS^c	0.83^c	0.01	0.20^c	0.02	354	7