Reduction strategies for hierarchical multi-label classification in protein function prediction

BMC Bioinformatics

Table 1 Summary of datasets: number of attributes (|A|), number of classes (|C|), number of classes per level (Classes per level), total number of instances (Total) and number of multi-label instances (Multi)

Dataset	\|A\|	\|C\|	Classes per level	Training		Valid		Test
				Total	Multi	Total	Multi	Total	Multi
1 - Seq [44]	478	499	18/80/178/142/77/4	1701	1344	879	679	1339	1079
2 - Pheno [44]	69	455	18/74/165/129/65/4	656	537	353	283	582	480
3 - Cellcycle [45]	77	499	18/80/178/142/77/4	1628	1323	848	673	1281	1059
4 - Church [46]	27	499	18/80/178/142/77/4	1630	1322	844	670	1281	1057
5 - Derisi [47]	63	499	18/80/178/142/77/4	1608	1309	842	671	1275	1055
6 - Eisen [48]	79	461	18/76/165/131/67/4	1058	900	529	441	837	719
7 - Expr [44]	551	499	18/80/178/142/77/4	1639	1328	849	674	1291	1064
8 - Gasch1 [49]	173	499	18/80/178/142/77/4	1634	1325	846	672	1284	1059
9 - Gasch2 [50]	52	499	18/80/178/142/77/4	1639	1328	849	674	1291	1064
10 - Spo [51]	80	499	18/80/178/142/77/4	1600	1301	837	666	1266	1047

ISSN: 1471-2105