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

Applied Network Science

Table 11 Louvain algorithm parameter optimization on generic data sets

Data set	Measure	Fold	Internal CV			External CV
Data set	Measure	Fold	Optimal resolution	AUC	AUPR	AUC	AUPR
MovieLens	JC\(_{CC}\)	1	0.3	0.79	0.18	0.79	0.23
		2	0.3	0.79	0.18	0.77	0.22
		3	0.3	0.80	0.19	0.76	0.23
		4	0.3	0.79	0.18	0.77	0.22
		5	0.3	0.79	0.18	0.77	0.22
	JC\(_{NC}\)	1	0.6	0.88	0.26	0.88	0.34
		2	0.8	0.88	0.27	0.88	0.35
		3	0.6	0.88	0.26	0.89	0.34
		4	0.7	0.88	0.26	0.88	0.33
		5	0.6	0.88	0.26	0.88	0.33
Unicodelang	JC\(_{CC}\)	1	0.6	0.73	0.10	0.72	0.08
		2	0.6	0.73	0.09	0.73	0.07
		3	0.6	0.73	0.09	0.74	0.10
		4	0.6	0.73	0.09	0.69	0.09
		5	0.6	0.73	0.10	0.73	0.15
	JC\(_{NC}\)	1	0.7	0.80	0.14	0.80	0.12
		2	0.7	0.79	0.13	0.80	0.13
		3	0.8	0.79	0.14	0.84	0.16
		4	0.7	0.81	0.11	0.79	0.16
		5	0.7	0.80	0.14	0.79	0.25