Transcriptome dynamics-based operon prediction in prokaryotes

BMC Bioinformatics

Table 4 Accuracy values from the 5-fold cross validation process, using all the features

Genome	Condition	Dataset	NNs	RFs	SVMs
H. somni	HS	Training set	0.95 (0.01)	0.97 (0.008)	0.97 (0.006)
		Test set	0.98	0.99	0.96
P. gingivalis	PG1	Training set	0.98 (0.002)	0.97 (0.004)	0.96 (0.004)
		Test set	0.99	0.98	0.95
	PG2	Training set	0.98 (0.003)	0.99 (0.004)	0.97 (0.003)
		Test set	0.98	0.98	0.98
	PG3	Training set	0.98 (0.001)	0.97 (0.002)	0.97 (0.002)
		Test set	0.98	0.98	0.90
E. coli	EC1	Training set	0.98 (0.009)	0.98 (0.003)	0.92 (0.008)
		Test set	0.98	0.99	0.95
	EC2	Training set	0.98 (0.003)	0.98 (0.006)	0.94 (0.009)
		Test set	0.99	0.99	0.99
S. enterica	SE1	Training set	0.95 (0.01)	0.97 (0.001)	0.91 (0.006)
		Test set	0.97	0.98	0.92
	SE2	Training set	0.94 (0.02)	0.98 (0.004)	0.90 (0.01)
		Test set	0.98	0.98	0.93

The accuracy values obtained with training (5-cross validation) and testing datasets are shown. The values are aggregates from all five folds.

ISSN: 1471-2105