Skip to main content

Table 5 Performance comparison of ten algorithms on six biological data sets, i.e. TCGA, BCWD, Dyrskjot-2003, Nutt-2003-v1, Nutt-2003-v3 and AIBT

From: Shrinkage Clustering: a fast and size-constrained clustering algorithm for biomedical applications

Data Metric Shrinkage Spectral K-means Hierarchical PAM DBSCAN Affinity AGNES Clusterdp SymNMF
TCGA NMI 0.91 0.77 NA 0.83 0.76 NA NA 0.82 NA 0.78
  Rand 0.97 0.91 NA 0.91 0.77 NA NA 0.90 NA 0.94
  F1 0.98 0.92 NA 0.92 0.80 NA NA 0.92 NA 0.95
  K (3) 3 2 NA 2 2 NA NA 3 NA 2
BCWD NMI 0.50 0.29 0.46 0.09 0.50 0.20 0.45 0.09 0.20 0.56
  Rand 0.77 0.68 0.75 0.55 0.77 0.64 0.76 0.55 0.53 0.83
  F1 0.80 0.69 0.79 0.69 0.80 0.75 0.79 0.69 0.59 0.85
  K (2) 2 2 2 2 2 2 3 2 2 2
Dyrskjot-2003 NMI 0.45 0.07 0.51 0.12 0.56 0.30 0.42 0.12 0.07 0.58
  Rand 0.78 0.55 0.76 0.42 0.77 0.55 0.72 0.42 0.50 0.83
  F1 0.70 0.36 0.71 0.54 0.66 0.60 0.66 0.54 0.43 0.75
  K (3) 3 3 3 3 3 3 3 3 2 3
Nutt-2003-v1 NMI 0.56 0.45 0.47 0.28 0.34 0.61 0.41 0.11 0.17 0.49
  Rand 0.72 0.73 0.72 0.52 0.68 0.65 0.73 0.35 0.64 0.72
  F1 0.58 0.51 0.51 0.43 0.41 0.62 0.44 0.38 0.34 0.55
  K (4) 4 4 4 4 4 4 5 4 4 4
Nutt-2003-v3 NMI 1.00 0.20 0.75 0.13 0.33 0.13 0.13 0.13 0.29 0.76
  Rand 1.00 0.58 0.91 0.58 0.58 0.58 0.58 0.58 0.55 0.91
  F1 1.00 0.59 0.92 0.71 0.60 0.71 0.71 0.71 0.57 0.91
  K (2) 2 2 2 2 2 2 3 2 2 2
AIBT NMI 0.56 0.20 0.58 0.17 0.54 0.56 0.53 0.02 0.55 0.55
  Rand 0.79 0.68 0.80 0.37 0.78 0.65 0.76 0.26 0.69 0.79
  F1 0.61 0.39 0.62 0.40 0.59 0.59 0.51 0.40 0.57 0.61
  K (4) 4 4 4 4 4 4 5 4 3 4
  1. Clustering accuracy is assessed via metrics including NMI (Normalized Mutual Information), Rand Index, F1 score and K (the optimal cluster number). The top three performers in each case are highlighted in bold