Table 2 Macro-F1 scores for classification of nodes in epidemic states according to different SIR epidemic processes over empirical datasets. For each \((\beta ,\mu )\) we highlight the two highest scores and underline the best one
From: Time-varying graph representation learning via higher-order skip-gram with negative sampling
(β,μ) | Model | Dataset | ||||
|---|---|---|---|---|---|---|
LyonSchool | SFHH | LH10 | Thiers13 | InVS15 | ||
(0.25,0.002) | DyANE | 78.1 ± 0.5 | 67.0 ± 1.2 | 52.5 ± 1.7 | 71.9 ± 0.6 | 64.3 ± 0.8 |
DynGEM | 58.7 ± 2.8 | 35.9 ± 1.1 | 34.5 ± 0.7 | 35.5 ± 1.2 | 58.8 ± 1.1 | |
DynamicTriad | 31.0 ± 0.4 | 28.8 ± 0.4 | 29.9 ± 0.3 | 30.3 ± 0.2 | 30.4 ± 0.2 | |
DySAT | 27.3 ± 0.2 | 27.4 ± 0.3 | 29.7 ± 0.2 | 30.2 ± 0.2 | 30.5 ± 0.2 | |
ISGNS | 63.5 ± 0.6 | 60.7 ± 0.8 | 54.1 ± 1.1 | 56.4 ± 0.6 | 52.3 ± 0.6 | |
\(\text{HOSGNS} ^{(\text{stat})}\) | 55.5 ± 0.8 | 57.3 ± 1.1 | 45.9 ± 0.9 | 46.9 ± 0.7 | 44.5 ± 0.7 | |
\(\text{HOSGNS} ^{(\text{dyn})}\) | \(\underline{\mathbf{79}\boldsymbol{.}\mathbf{2}}\pm \mathbf{0}\boldsymbol{.}\mathbf{5}\) | \(\underline{\mathbf{69}\boldsymbol{.}\mathbf{1}}\pm \mathbf{1}\boldsymbol{.}\mathbf{1}\) | 59.6 ± 1.5 | 71.8 ± 1.2 | \(\underline{\mathbf{64}\boldsymbol{.}\mathbf{6}}\pm \mathbf{0}\boldsymbol{.}\mathbf{7}\) | |
\(\text{HOSGNS} ^{(\text{stat}|\text{dyn})}\) | 77.4 ± 0.6 | 67.4 ± 1.2 | \(\underline{\mathbf{59}\boldsymbol{.}\mathbf{7}}\pm \mathbf{1}\boldsymbol{.}\mathbf{2}\) | \(\underline{\mathbf{72}\boldsymbol{.}\mathbf{5}}\pm \mathbf{0}\boldsymbol{.}\mathbf{7}\) | 64.2 ± 1.0 | |
(0.0625,0.002) | DyANE | 72.2 ± 0.6 | 64.9 ± 1.7 | 59.0 ± 1.2 | 68.0 ± 0.5 | \(\underline{\mathbf{60}\boldsymbol{.}\mathbf{2}}\pm \mathbf{0}\boldsymbol{.}\mathbf{5}\) |
DynGEM | 56.4 ± 2.7 | 35.9 ± 4.1 | 35.8 ± 1.2 | 32.9 ± 1.2 | 55.0 ± 0.6 | |
DynamicTriad | 29.5 ± 0.5 | 33.1 ± 2.5 | 29.6 ± 0.4 | 27.4 ± 0.3 | 28.4 ± 0.2 | |
DySAT | 26.4 ± 0.2 | 29.5 ± 1.3 | 29.5 ± 0.3 | 26.5 ± 0.2 | 28.5 ± 0.2 | |
ISGNS | 59.2 ± 0.3 | 57.1 ± 1.6 | 55.9 ± 1.0 | 49.0 ± 0.3 | 47.2 ± 0.3 | |
\(\text{HOSGNS} ^{(\text{stat})}\) | 55.5 ± 0.7 | 57.6 ± 2.2 | 49.4 ± 0.8 | 45.5 ± 0.4 | 43.6 ± 0.5 | |
\(\text{HOSGNS} ^{(\text{dyn})}\) | \(\underline{\mathbf{73}\boldsymbol{.}\mathbf{5}}\pm \mathbf{0}\boldsymbol{.}\mathbf{5}\) | 65.7 ± 1.6 | \(\underline{\mathbf{61}\boldsymbol{.}\mathbf{1}}\pm \mathbf{1}\boldsymbol{.}\mathbf{2}\) | \(\underline{\mathbf{69}\boldsymbol{.}\mathbf{5}}\pm \mathbf{0}\boldsymbol{.}\mathbf{3}\) | 59.6 ± 0.5 | |
\(\text{HOSGNS} ^{(\text{stat}|\text{dyn})}\) | 72.9 ± 0.6 | \(\underline{\mathbf{66}\boldsymbol{.}\mathbf{3}}\pm \mathbf{1}\boldsymbol{.}\mathbf{9}\) | 58.2 ± 1.1 | 68.5 ± 0.4 | 59.0 ± 0.7 | |
(0.1875,0.001) | DyANE | 74.7 ± 0.7 | 67.7 ± 1.2 | \(\underline{\mathbf{63}\boldsymbol{.}\mathbf{4}}\pm \mathbf{1}\boldsymbol{.}\mathbf{8}\) | 72.7 ± 0.4 | \(\underline{\mathbf{68}\boldsymbol{.}\mathbf{6}}\pm \mathbf{0}\boldsymbol{.}\mathbf{4}\) |
DynGEM | 57.4 ± 2.8 | 36.2 ± 2.6 | 41.4 ± 1.3 | 34.8 ± 1.3 | 61.2 ± 0.9 | |
DynamicTriad | 32.3 ± 0.5 | 31.5 ± 0.8 | 30.5 ± 0.4 | 27.9 ± 0.3 | 30.0 ± 0.2 | |
DySAT | 26.4 ± 0.2 | 29.4 ± 0.8 | 30.0 ± 0.3 | 27.7 ± 0.3 | 29.9 ± 0.2 | |
ISGNS | 65.1 ± 0.5 | 63.0 ± 1.4 | 60.2 ± 1.7 | 56.0 ± 0.5 | 52.5 ± 0.5 | |
\(\text{HOSGNS} ^{(\text{stat})}\) | 56.9 ± 0.8 | 59.4 ± 1.7 | 48.5 ± 1.1 | 49.0 ± 0.6 | 46.2 ± 0.8 | |
\(\text{HOSGNS} ^{(\text{dyn})}\) | \(\underline{\mathbf{76}\boldsymbol{.}\mathbf{5}}\pm \mathbf{0}\boldsymbol{.}\mathbf{4}\) | 68.6 ± 1.1 | 62.4 ± 1.7 | \(\underline{\mathbf{74}\boldsymbol{.}\mathbf{8}}\pm \mathbf{0}\boldsymbol{.}\mathbf{5}\) | 67.9 ± 0.7 | |
\(\text{HOSGNS} ^{(\text{stat}|\text{dyn})}\) | 74.5 ± 0.4 | \(\underline{\mathbf{69}\boldsymbol{.}\mathbf{4}}\pm \mathbf{1}\boldsymbol{.}\mathbf{4}\) | 62.5 ± 2.0 | 73.6 ± 0.6 | 67.3 ± 0.5 | |