Skip to main content
. 2018 Mar 16;9:476. doi: 10.3389/fmicb.2018.00476

Figure 3.

Figure 3

(A) Performance of SVM-based classifiers in distinguishing between PVPs and non-PVPs. A total of 25 classifiers were evaluated using LOOCV, and their performances in terms of MCC, accuracy, sensitivity, and specificity are shown. The red arrow denotes the final selected model. (B) Classification accuracy of the F136-based (final selected) model with respect to variations in parameters C and γ. (C) Distribution of each feature type in the optimal feature set (136 features) and original feature set (583 features).