Table 4 Macro-F1 scores for missing event prediction in empirical datasets. We highlight in bold the two best scores for each dataset. For baseline models we underline their highest score
From: Time-varying graph representation learning via higher-order skip-gram with negative sampling
Model | Operator | Dataset | ||||
|---|---|---|---|---|---|---|
LyonSchool | SFHH | LH10 | Thiers13 | InVS15 | ||
DyANE | Average | 56.8 ± 0.6 | 50.6 ± 0.8 | 51.3 ± 1.0 | 49.1 ± 0.6 | 49.3 ± 0.8 |
Hadamard | 87.3 ± 0.3 | 73.5 ± 0.6 | \(\underline{67.0}\pm 1.0\) | \(\underline{87.2}\pm 0.3\) | \(\underline{80.1}\pm 0.8\) | |
Weighted-L1 | 87.8 ± 0.3 | 73.3 ± 0.6 | 65.9 ± 1.0 | 84.0 ± 0.4 | 78.4 ± 0.6 | |
Weighted-L2 | \(\underline{88.5}\pm 0.2\) | \(\underline{73.7}\pm 0.5\) | 66.1 ± 1.0 | 84.4 ± 0.4 | 78.9 ± 0.6 | |
Concat | 64.4 ± 0.5 | 52.4 ± 0.8 | 51.9 ± 1.0 | 57.0 ± 0.6 | 51.4 ± 0.7 | |
DynGEM | Average | 56.2 ± 0.5 | \(\underline{51.8}\pm 0.8\) | \(\underline{52.0}\pm 1.1\) | 49.7 ± 0.5 | 50.9 ± 0.7 |
Hadamard | 54.8 ± 0.6 | 51.3 ± 0.7 | 51.7 ± 1.2 | 44.7 ± 0.7 | \(\underline{50.9}\pm 0.6\) | |
Weighted-L1 | 55.5 ± 0.4 | 48.5 ± 0.8 | 50.2 ± 1.0 | \(\underline{52.2}\pm 0.4\) | 49.8 ± 0.7 | |
Weighted-L2 | 53.2 ± 0.7 | 47.8 ± 0.9 | 48.0 ± 1.1 | 48.9 ± 0.6 | 45.3 ± 0.6 | |
Concat | \(\underline{58.2}\pm 0.5\) | 50.4 ± 0.8 | 46.4 ± 1.4 | 48.8 ± 0.5 | 49.9 ± 0.6 | |
DynamicTriad | Average | 51.4 ± 0.4 | 52.6 ± 0.6 | 53.0 ± 0.8 | 52.0 ± 0.4 | 49.9 ± 0.7 |
Hadamard | 53.1 ± 0.4 | 49.5 ± 0.6 | 52.0 ± 0.8 | 51.7 ± 0.5 | 49.8 ± 0.6 | |
Weighted-L1 | 64.3 ± 0.4 | 56.6 ± 0.7 | 54.2 ± 0.9 | 53.6 ± 0.4 | 47.2 ± 0.6 | |
Weighted-L2 | \(\underline{64.5}\pm 0.4\) | \(\underline{57.3}\pm 0.7\) | \(\underline{54.9}\pm 0.9\) | \(\underline{54.5}\pm 0.5\) | 47.0 ± 0.6 | |
Concat | 52.6 ± 0.3 | 51.8 ± 0.5 | 52.7 ± 0.9 | 51.5 ± 0.3 | \(\underline{49.9}\pm 0.6\) | |
DySAT | Average | 51.3 ± 0.4 | 51.6 ± 0.6 | 52.5 ± 0.8 | 50.0 ± 0.4 | 50.3 ± 0.6 |
Hadamard | \(\underline{73.8}\pm 0.6\) | \(\underline{52.5}\pm 0.7\) | 56.6 ± 0.7 | \(\underline{68.5}\pm 0.5\) | 61.5 ± 0.8 | |
Weighted-L1 | 71.3 ± 0.5 | 52.0 ± 0.6 | \(\underline{57.6}\pm 0.8\) | 63.2 ± 0.6 | \(\underline{64.4}\pm 0.5\) | |
Weighted-L2 | 70.7 ± 0.5 | 51.5 ± 0.7 | 56.5 ± 0.8 | 63.1 ± 0.5 | 63.4 ± 0.5 | |
Concat | 49.2 ± 0.4 | 48.8 ± 0.8 | 52.4 ± 0.9 | 49.8 ± 0.5 | 50.4 ± 0.6 | |
ISGNS | Average | 52.4 ± 0.6 | 49.5 ± 0.8 | 44.9 ± 0.9 | 48.0 ± 0.4 | 42.7 ± 0.8 |
Hadamard | 79.8 ± 0.4 | 59.3 ± 0.7 | 61.1 ± 1.2 | 59.3 ± 0.6 | \(\underline{51.7}\pm 0.7\) | |
Weighted-L1 | 80.8 ± 0.3 | 59.8 ± 0.7 | 61.7 ± 1.0 | 59.0 ± 0.6 | 49.8 ± 0.7 | |
Weighted-L2 | \(\underline{81.5}\pm 0.3\) | \(\underline{60.2}\pm 0.7\) | \(\underline{62.5}\pm 0.9\) | \(\underline{59.9}\pm 0.6\) | 51.5 ± 0.7 | |
Concat | 55.8 ± 0.7 | 50.8 ± 0.6 | 46.8 ± 0.8 | 52.2 ± 0.5 | 48.5 ± 0.6 | |
\(\text{HOSGNS} ^{(\text{stat})}\) | Hadamard | 52.1 ± 0.4 | 43.8 ± 0.6 | 34.2 ± 0.2 | 55.9 ± 0.6 | 43.0 ± 0.5 |
\(\text{HOSGNS} ^{(\text{dyn})}\) | Hadamard | 89.2 ± 0.2 | 74.9 ± 0.6 | 67.1 ± 0.8 | 90.7 ± 0.3 | 81.4 ± 0.5 |
\(\text{HOSGNS} ^{(\text{stat}|\text{dyn})}\) | Hadamard | 89.2 ± 0.3 | 76.3 ± 0.7 | 68.5 ± 1.0 | 89.9 ± 0.3 | 80.8 ± 0.6 |