Table 3.

Comparison of models learned using bound selection (CART_b, SCM_b) and cross-validation (CART_cv, SCM_cv) as model selection strategies.

Dataset	SCMb	SCMcv	CARTb	CARTcv
A. baumannii	0.849 (2.7)	0.857 (10.6)	0.864 (3.4)	0.863 (9.6)
E. coli	0.818 (4.6)	0.830 (6.2)	0.808 (7.0)	0.812 (13.3)
E. faecium	1.000 (1.0)	1.000 (1.0)	1.000 (1.0)	1.000 (1.0)
K. pneumoniae	0.950 (3.9)	0.953 (7.9)	0.949 (4.3)	0.948 (8.8)
M. tuberculosis	0.963 (4.5)	0.963 (5.0)	0.962 (4.7)	0.963 (5.9)
N. gonorrhoeae	0.935 (3.0)	0.935 (3.5)	0.936 (3.3)	0.929 (6.1)
P. aeruginosa	0.939 (1.2)	0.939 (1.4)	0.942 (1.1)	0.941 (2.5)
P. difficile	0.982 (1.0)	0.982 (1.0)	0.982 (1.0)	0.982 (1.1)
S. aureus	0.987 (1.0)	0.987 (1.9)	0.987 (1.0)	0.987 (1.6)
S. enterica	0.913 (1.0)	0.907 (1.6)	0.913 (1.0)	0.900 (3.6)
S. haemolyticus	0.925 (1.0)	0.933 (1.0)	0.925 (1.0)	0.933 (1.0)
S. pneumoniae	0.960 (1.0)	0.959 (1.0)	0.960 (1.0)	0.959 (1.0)

For each dataset the accuracy is shown, along with the number of k-mers used by the model (in parentheses). Accuracies within 1% of the maximum value are shown in bold.