Research Article

Semi-Supervised Predictive Clustering Trees for (Hierarchical) Multi-Label Classification

Table 5

Model sizes expressed as the number of nodes in trees.

DatasetNumber of labeled examples
50100200350500
SL-PCTSSL-PCTSL-PCTSSL-PCTSL-PCTSSL-PCTSL-PCTSSL-PCTSL-PCTSSL-PCT

Multi-label classification
Bibtex121121.4119.4119.4119
Birds115.8120.81.215.22.613412.8
Corel5k133.611111111
Emotions318.85197.47.411.819.214.814.8
Enron117.6119.2121.4121.21.220.2
Genbase2.622.821.42331.226.637.236.84341.4
Mediana1.432.8447.87.810.210.212.412.4
Medical115.4141.8143.43.263.613.663.2
Scene7.825.812.428.419.828.431.82936.637.2
SIGMEA real335.23.23.24.64.66.66.68.48.4
Slovenian rivers139.21641.467.6370.83.855.8
Yeast11111251.2252.225
Average2.123.34.420.66.522.39.226.311.825.9

Hierarchical multi-label classification
Danish farms111.41.43.23.2668.8257.6
Slovenian rivers119.8118.2147.41.452.4350.8
Enron111.6113.81.615.85.616.66.617.2
ImCLEF07A142.8286.66.6133.412.8150.419.4177.4
ImCLEF07D147.41.82154.4295.27159.215172.2
Diatoms149.6156.2157.4172.4189.8
Cellcycle-GO1111111111
Church-GO1111111111
Derisi-GO1111111111
Eisen-GO11111111127.4
Expr-GO11111119.219.2
Pheno-GO1111111111
Average1.014.91.233.12.046.53.339.35.067.1