Diagnosis of ventilator-associated pneumonia using electronic nose sensor array signals: solutions to improve the application of machine learning in respiratory research

Table 2 Prediction accuracy of the electronic nose in the test set of machine learning algorithms

Model and parameters	Accuracy (95% CI)	Sensitivity	Specificity	PPV	NPV	Kappa	AUC (95% CI)
k-nearest neighbors (k = 5)	0.77 (0.46–0.95)	0.71	0.83	0.83	0.71	0.54	0.80 (0.54–1.00)
Naive Bayes (fL = 0, usekernel = TRUE, adjust = 1)	0.77 (0.46–0.95)	0.71	0.83	0.83	0.71	0.54	0.80 (0.54–1.00)
Decision tree (trials = 10, model = rules, window = TRUE)	0.85 (0.55–0.98)	0.86	0.83	0.86	0.83	0.69	0.85 (0.63–1.00)
Neural network (size = 3, decay = 1e-04)	0.85 (0.55–0.98)	0.86	0.83	0.86	0.83	0.69	0.85 (0.63–1.00)
Support vector machines (linear kernel) (C = 1)	0.85 (0.55–0.98)	0.86	0.83	0.86	0.83	0.69	0.85 (0.63–1.00)
Support vector machines (radial kernel) (sigma = 1.432815, C = 1)	0.77 (0.46–0.95)	0.71	0.83	0.83	0.71	0.54	0.85 (0.63–1.00)
Support vector machines (polynomial kernel) (degree = 1, scale = 0.1, C = 0.5)	0.85 (0.55–0.98)	0.86	0.83	0.86	0.83	0.69	0.85 (0.63–1.00)
Random forest (mtry = 32)	0.77 (0.46–0.95)	0.71	0.83	0.83	0.71	0.54	0.90 (0.74–1.00)
Mean value (SD)	0.81 (0.04)	0.79 (0.08)	0.83 (0.00)	0.85 (0.02)	0.77 (0.06)	0.62 (0.08)	0.85 (0.04)

PPV positive predictive value; NPV negative predictive value; AUC area under the receiver operating curve