Co-MLHAN: contrastive learning for multilayer heterogeneous attributed networks

Applied Network Science

Table 6 Results on IMDb-MLH (4000 features), with early stopping

Method	F1 micro	F1 macro	AUC	F1 ‘action’	F1 ‘comedy’	F1 ‘drama’
Co-MLHAN	0.6801 ± 0.0111	0.6308 ± 0.0386	0.7968 ± 0.0154	0.5017 ± 0.1106	0.6986 ± 0.0093	0.6922 ± 0.0104
Co-MLHAN-SA	0.6555 ± 0.0149	0.6124 ± 0.0407	0.7788 ± 0.0207	0.4996 ± 0.1132	0.6652 ± 0.0142	0.6724 ± 0.0132
HeCo	0.5053 ± 0.0044	0.3447 ± 0.0060	0.5682 ± 0.0082	0.2598 ± 0.0232	0.6106 ± 0.0144	0.6569 ± 0.0139
NSHE	0.6052 ± 0.0120	0.5091 ± 0.0115	0.7020 ± 0.0106	0.0014 ± 0.0097	0.5931 ± 0.0046	0.4397 ± 0.0081