Table 4

Influence of the negative:positive training set ratio on the prediction accuracy

Neg:pos testing ratio

1:1

1:1

100:1

100:1

ROC50 AUC (std)^a

ROC100 AUC (std)^a

ROC50 AUC (std)^a

ROC100 AUC (std)^a

Neg:pos training ratio

1:1

0.300 (0.008)

0.385 (0.006)

0.0645 (0.0019)

0.0814 (0.0009)

100:1

0.325 (0.004)

0.403 (0.003)

0.0747 (0.0022)

0.0944 (0.0028)

^a The ROCn AUCs are an average of five separate experiments (each of which is itself a five-fold cross validation experiment). Their standard deviation is shown in parenthesis.

Scott and Barton BMC Bioinformatics 2007 8:239   doi:10.1186/1471-2105-8-239