MScanner: a classifier for retrieving Medline citations

BMC Bioinformatics

Table 2 Cross validation statistics.

Statistic	PG07	Radiology	AIDSBio	Control
# Relevant	1663	67	10727	10000
# Irrelevant	99986	100000	99927	99955
Prevalence	0.01636	0.00067	0.09702	0.09095
ROC Area	0.9754	0.9923	0.9913	0.4975
ROC Std Error	0.0020	0.0047	0.0004	0.0030
Averaged Precision	0.693	0.711	0.924	0.090
Break-Even	0.652	0.642	0.884	0.089

Summary of the cross validation training sets and performance metrics. Prevalence is the fraction of the data that is relevant, and break-even is point where cross validation precision equals recall.

ISSN: 1471-2105