Bayesian network models with decision tree analysis for management of childhood malaria in Malawi

Table 3 Performance of the models computed using tenfold cross-validation

Performance metric	Threshold	Manual BN	TAN	Logistic regression	Random forest
Area under ROC curve (AUC)	–	0.60	0.58	0.57	0.57
Balanced accuracy (BAC)	Default	0.56	0.53	0.50	0.50
Sensitivity	Default	0.32	0.22	0.07	0.03
Specificity	Default	0.80	0.84	0.93	0.97
Balanced accuracy (BAC)	Optimal	0.58	0.57	0.57	0.57
Sensitivity	Optimal	0.74	0.45	0.59	0.52
Specificity	Optimal	0.42	0.68	0.55	0.62

ISSN: 1472-6947