Machine learning methods for metabolic pathway prediction

BMC Bioinformatics

Table 6 Naïve Bayes performance

Predictor	AUC	max. ACC	SN (max. ACC)	SP (max. ACC)	max. FM	PR (max. FM)	RC (max. FM)
all features	0.91	0.883	0.763	0.915	0.736	0.68	0.804
random features (r = 37)	0.916	0.884	0.686	0.935	0.725	0.67	0.792
random forest (r = 37, c = 60)	0.924	0.888	0.709	0.936	0.737	0.693	0.791
HC-BIC feature selection	0.933	0.905	0.787	0.936	0.775	0.757	0.794
HC-AIC feature selection	0.938	0.905	0.78	0.938	0.777	0.759	0.796
bagged HC-BIC (c = 15)	0.945	0.908	0.751	0.949	0.782	0.761	0.805
bagged HC-AIC (c = 15)	0.946	0.909	0.757	0.949	0.78	0.767	0.796

See Table 3 for description of column headings. HC-BIC = hill-climbing on Bayes information criterion; HC-AIC = hill-climbing on Akaike information criterion.

ISSN: 1471-2105