Large-scale online semantic indexing of biomedical articles via an ensemble of multi-label classification models

Table 6 Comparison of the three ensemble methods regarding the number of labels predicted by each model

# of labels predicted from each model
	MetaLabeler	SVM _Tuned	SVM _Vanilla	LLDA
Data set A
Improve micro-F	10751	15002
Improve F [13]	11256	14497
MULE	25192	561
Improve micro-F	19549			6204
Improve F [13]	15293			10460
MULE	25322			431
Improve micro-F			18862	6891
Improve F [13]			12900	12853
MULE			25702	51
Improve micro-F		8213	17037	503
Improve F [13]		8723	16351	679
MULE		25210	526	17
Improve micro-F	10066	2938	2499	250
Improve F [13]	10887	2815	11782	269
MULE	24814	174	760	5
Data set B
Improve micro-F	4252	12059
Improve F [13]	4699	11612
MULE	16053	258
Improve micro-F	9342			6969
Improve F [13]	10920			5391
MULE	15826			485
Improve micro-F			1500	14811
Improve F [13]			801	15510
MULE			15998	313
Improve micro-F		1804	12774	1733
Improve F [13]		1732	12688	1891
MULE		16121	38	152
Improve micro-F	3817	494	11331	669
Improve F [13]	4198	400	11053	660
MULE	15736	144	117	43

The numbers are given for the micro-F optimization (first series of experiments)

ISSN: 2041-1480