Table 8 Results on IMDb-MLH-mb (1000 features) and positives selection AL1A
From: Co-MLHAN: contrastive learning for multilayer heterogeneous attributed networks
Method | Early-stopping | F1 micro | F1 macro | AUC |
|---|---|---|---|---|
Co-MLHAN (EL) | No | 0.7387 ± 0.0092 | 0.7419 ± 0.0087 | 0.8649 ± 0.0056 |
Co-MLHAN-SA (EL) | No | 0.7373 ± 0.0108 | 0.7469 ± 0.0102 | 0.8717 ± 0.0055 |
Co-MLHAN (NL) | No | 0.8611 ± 0.0084 | 0.8588 ± 0.0081 | 0.9437 ± 0.0040 |
Co-MLHAN-SA (NL) | No | 0.8676 ± 0.0067 | 0.8658 ± 0.0065 | 0.9449 ± 0.0036 |
Co-MLHAN (EL) | Yes | 0.7442 ± 0.0070 | 0.7480 ± 0.0070 | 0.8681 ± 0.0051 |
Co-MLHAN-SA (EL) | Yes | 0.7449 ± 0.0075 | 0.7529 ± 0.0076 | 0.8808 ± 0.0042 |
Co-MLHAN (NL) | Yes | 0.8482 ± 0.0138 | 0.8478 ± 0.0130 | 0.9389 ± 0.0066 |
Co-MLHAN-SA (NL) | Yes | 0.8678 ± 0.0066 | 0.8658 ± 0.0072 | 0.9517 ± 0.0031 |