Predicting Mycobacterium tuberculosis Complex Clades Using Knowledge-Based Bayesian Networks

Table 4

Average 𝐹-measure of KBBN, BN, Rules-only, and SVM (nonlinear and linear) on two testbeds. While using Rules-only provides poor results, KBBN is able to provide results that are significantly better or at least not worse than BN and SVM on both domains. Results significantly different from KBBN at 5% significance level are shown in bold.

Dataset	Model
	KBBN	BN	Rules-only	SVM nonlinear	SVM linear
SITVIT-CV	0.945	0.771	0.345	0.903	0.914
CDC-Sublineage	0.981	0.934	0.312	0.994	0.993