Predicting drug side effects by multi-label learning and ensemble learning

BMC Bioinformatics

Table 3 The 5-CV performances of individual feature-based FS-MLKNN models on Liu’s dataset

Features	AUC	AUPR	Hamming loss	Ranking loss	One error	Coverage	Average precision
Enzyme	0.8878 ± 0.0004	0.4080 ± 0.0013	0.0478 ± 0.0001	0.0826 ± 0.0002	0.1611 ± 0.0057	837.1250 ± 2.9063	0.4652 ± 0.0005
Pathway	0.8895 ± 0.0006	0.4187 ± 0.0028	0.0473 ± 0.0001	0.0792 ± 0.0003	0.1688 ± 0.0037	824.2678 ± 4.2341	0.4799 ± 0.0006
Target	0.8962 ± 0.0007	0.4557 ± 0.0019	0.0457 ± 0.0001	0.0739 ± 0.0003	0.1442 ± 0.0048	810.4788 ± 2.9801	0.5008 ± 0.0008
Transporter	0.8871 ± 0.0008	0.4060 ± 0.0018	0.0480 ± 0.0001	0.0819 ± 0.0003	0.1635 ± 0.0037	836.4404 ± 2.3029	0.4698 ± 0.0007
Indication	0.8963 ± 0.0008	0.4648 ± 0.0043	0.0452 ± 0.0002	0.0755 ± 0.0003	0.1341 ± 0.0054	818.0483 ± 3.9917	0.5005 ± 0.0014
Substructure	0.8931 ± 0.0005	0.4343 ± 0.0011	0.0468 ± 0.0001	0.0739 ± 0.0005	0.1659 ± 0.0069	804.3813 ± 2.7354	0.4989 ± 0.0021

ISSN: 1471-2105