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

BMC Bioinformatics

Table 6 Classification results on the 190 PD pairs with CV evaluation

Kernel	r	TP	e	TP/e	TP/#PD
SpT	46.4	71	88	0.81	0.37
PT	27.9	33	53	0.62	0.17
kBSPS	36.6	22	70	0.31	0.12
ST	28.0	19	53	0.36	0.10
SST	26.6	16	51	0.31	0.08
APG	25.3	15	48	0.31	0.08
SL	30.4	15	58	0.26	0.08
syntactic	24.4	14	46	0.30	0.07
edit	18.1	11	34	0.32	0.06
lexical	25.0	9	47	0.19	0.05
shallow	24.6	7	47	0.15	0.04
cosine	24.9	7	47	0.15	0.04
combined	24.8	4	47	0.09	0.02

Classification results on the 190 PD 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; TP is the number of correctly classified PD pairs; e is 190·r, the expected number of negative class predictions projected onto the 190 PD pairs.

ISSN: 1471-2105