Table 3. Summary of the diagnostic accuracy of the machine learning algorithm analysis.

Algorithm	Classifier type	Entire sample set			Africa samples
		Sensitivity	Specificity	ROC area	Sensitivity	Specificity	ROC area
BayesNet	Bayesian network	84.8	87	0.929	97.3	99.1	1
NaiveBayes	Bayesian network	88.6	88.3	0.930	94.5	98.2	1
Logistic	Logistic regression	85.6	85.2	0.898	98.6	99.6	0.999
IB1	Nearest neighbor	87.1	83.9	0.855	98.6	95.8	0.972
OneR	Minimum error attribute	76.5	76.6	0.766	89.0	85.2	0.871
Multilayer perceptron	Backpropagation classification	87.9	85.9	0.921	98.6	95.8	1
FLR	Fuzzy lattice reasoning	81.1	83.7	0.824	98.6	99.6	0.991
Functional trees	Classification tree	84.1	83.6	0.861	100	100	1
Random forest	Classification tree	88.6	89.3	0.954	97.3	95.4	0.997

The mass spectral ion intensities of the top 14 candidate onchocerciasis biomarkers from onchocerciasis infected and uninfected samples were compared between the multi-region sample set and the African blood samples. All results were obtained using a 10 fold cross validation analysis.