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

Table 3

HMLC datasets and their characteristics.

Dataset	Domain
Danish farms [42]	Ecology	1944	132/5	Tree	72	3	7
Enron [34]	Text	1702	1001/0	Tree	53	2	3.38
Slovenian rivers [40]	Ecology	1060	0/16	Tree	724	4	25
ImCLEF07A [43]	Images	11006	0/80	Tree	96	3	1
ImCLEF07D [43]	Images	11006	0/80	Tree	46	3	1
Diatoms [44]	Images	3119	0/371	Tree	377	3	0.94
Cellcycle-GO [25]	Genomics	3766	0/77	DAG	4126	12	35.91
Church-GO [25]	Genomics	3764	1/26	DAG	4126	12	35.89
Derisi-GO [25]	Genomics	3733	0/63	DAG	4120	12	35.99
Eisen-GO [25]	Genomics	2425	0/79	DAG	3574	11.12	39.04
Expr-GO [25]	Genomics	3788	4/547	DAG	4132	12	35.87
Pheno-GO [25]	Genomics	1592	69/0	DAG	3128	12	36.43

is the number of examples, is the number of descriptive variables (nominal/continuous), is the type of the label hierarchy, is the number of nodes in the hierarchy, is the maximal depth of the hierarchy, and is the average number of labels per example.