A detailed error analysis of 13 kernel methods for protein-protein interaction extraction

BMC Bioinformatics

Table 4 Classification results on the 521 ND pairs with CV evaluation

Kernel	r	TN	e	TN/e	TN/ND
edit	18.1	305	427	0.71	0.59
lexical	25.0	203	391	0.52	0.39
SST	26.6	186	382	0.49	0.36
APG	25.3	185	389	0.48	0.36
PT	27.9	185	376	0.49	0.36
syntactic	24.4	180	394	0.46	0.35
cosine	24.9	168	391	0.43	0.32
ST	28.0	160	375	0.43	0.30
shallow	24.6	136	393	0.35	0.26
kBSPS	36.6	122	330	0.37	0.23
combined	24.8	117	392	0.30	0.22
SL	30.4	116	363	0.32	0.22
SpT	46.4	88	279	0.32	0.17

Classification results on the 521 ND pairs with CV evaluation (in decreasing order according to the number of successfully classified pairs). Ratio (r) refers to the distribution of positive classes predicted by the kernel measured across the 5 corpora; TN is the number of correctly classified ND pairs; e is 521·(1−r), the expected number of negative class predictions projected onto the 521 ND pairs.

ISSN: 1471-2105