Skip to main content

Table 4 Comparison of clustering methods on two data sets

From: A clustering-based approach for efficient identification of microRNA combinatorial biomarkers

Feature combination Performance measure GSE22220 GSE40525
   HCb RCa HCb RCa
Pair AvgRank 10 7 8.2 8.5 9.1
  AvgRank 100 92.2 84.4 180.9 103.0
  AvgRank 1000 2003.1 1624.4 10696.8 1470.5
  HitRatio 10(%) 70.0 70.0 60.0 40.0
  HitRatio 100(%) 54.0 58.0 32.0 64.0
  HitRatio 1000(%) 30.0 32.0 15.0 39.2
Triple AvgRank 10 8.2 8.6 9.2 26.3
  AvgRank 100 95.4 94.3 68.7 229.4
  AvgRank 1000 1776.2 1684.3 3675.2 1577.2
  HitRatio 10(%) 60.0 60.0 20.0 20.0
  HitRatio 100(%) 58.0 58.0 71.0 66.0
  HitRatio 1000(%) 27.1 27.1 30.3 36.8
Quadruple AvgRank 10 14.2 12.8 9.0 174.0
  AvgRank 100 112.6 108.5 257.2 273.0
  AvgRank 1000 1639.1 1482.3 3171.3 2826.6
  HitRatio 10(%) 30.0 40.0 78.0 40.0
  HitRatio 100(%) 48.0 50.0 12.0 4.0
  HitRatio 1000(%) 23.2 26.4 16.0 19.3
  1. aRC: refined clustering, in which the inconsistency coefficient for raw clusters and thresholds of MSL and MLR are fixed
  2. bHC: hierarchical clustering, which performs the best by trying different inconsistency coefficients