ROC-Boosting: A Feature Selection Method for Health Identification Using Tongue Image

Figure 5

Two features of the AUC value larger than 0.5 are shown. Feature 181,520 in the left is selected, because the whole ROC curve is laid upward the random guess line. Feature 188,479 in the right is excluded, because its ROC curve crosses the random guess line at the corners of the ROC space.

(a)

(b)