Skip to main content

Table 4 The best clustering performance on two datasets

From: MHSNMF: multi-view hessian regularization based symmetric nonnegative matrix factorization for microbiome data analysis

 

Accuracy (%)

NMI (%)

Three-source

HMP

Three-source

HMP

BSSV

79.88

88.54

69.66

84.64

WSSV

65.68

81.16

58.26

80.71

Multi-NMF

66.86

77.55

55.04

72.87

Co-training SC

61.54

63.58

58.03

63.68

SNF

65.68

91.21

56.34

89.20

LJ-NMF

69.82

73.16

60.08

67.77

CSMF

65.18

74.01

63.23

65.43

NetNMF

70.18

82.50

61.24

81.76

MHSNMF

82.84

95.28

71.43

91.76

  1. In Multi-NMF, these clustering results on three-source and HMP data are obtained when γv = 0.01 and 0.05, respectively. For three-source dataset, the cosine function was used to construct the similarity matrix. For BSSV, WSSV and LJ-NMF, the number of neighborhoods on HMP data was set to be 12. For other values, MHSNMF still outperforms other algorithms in most cases.