Skip to main content

Table 2 Rand Index, Purity and Average F1 scores against ground truth; the number of clusters for hMETIS, PaToH, Spectral, and Zhou-Spectral is set to the number of clusters returned by the IRMM method are 13, 79, 8, 18, and 1358 for Citeseer, Cora, Movielens, TwitterFootball, and Arnetminer, respectively

From: Hypergraph clustering by iteratively reweighted modularity maximization

  Citeseer Cora MovieLens TwitterFootball Arnetminer
(a) Rand Index scores against ground truth.
hMETIS 0.6504 0.7592 0.4970 0.7639 0.0416
PaToH 0.6612 0.6919 0.4987 0.7553 0.0052
Spectral 0.7164 0.2478 0.4806 0.7486 0.0610
Zhou-Spectral 0.8210 0.5743 0.4977 0.9016 0.0628
Louvain 0.7361 0.7096 0.4898 0.6337 0.0384
NDP-Louvain 0.7899 0.8238 0.4988 0.9056 0.0821
IRMM 0.7986 0.8646 0.5091 0.9448 0.0967
(b) Purity scores against ground truth.
hMETIS 0.5894 0.6596 0.6893 0.2556 0.6831
PaToH 0.6271 0.5912 0.7017 0.3176 0.3928
Spectral 0.4629 0.3897 0.6832 0.8114 0.9216
Zhou-Spectral 0.5287 0.4145 0.7118 0.8325 0.9378
Louvain 0.7190 0.6836 0.7189 0.8054 0.9138
NDP-Louvain 0.7307 0.7597 0.7245 0.8829 0.9691
IRMM 0.7659 0.8138 0.7291 0.8948 0.9765
(c)Average F1 scores against ground truth.
hMETIS 0.1087 0.1075 0.1291 0.3197 0.0871
PaToH 0.0532 0.1171 0.1104 0.1132 0.0729
Spectral 0.1852 0.1291 0.1097 0.4496 0.0629
Zhou-Spectral 0.2774 0.2517 0.118 0.5055 0.0938
Louvain 0.1479 0.2725 0.1392 0.2238 0.1378
NDP-Louvain 0.2782 0.3248 0.1447 0.5461 0.1730
IRMM 0.4019 0.3709 0.1963 0.5924 0.1768
  1. Louvain, NDP-Louvain, and IRMM return the number of clusters on their own
  2. Best performance in each column is boldfaced