onstructed using the sigmoid kernel function with the optimal

rameters.

(a) (b)

The ROC curves of a SVM models for the breast cancer data. (a) Employing the

l kernel. The AUC was 0.994. (b) Employing the sigmoid kernel. The AUC

known that peptides are composed of non-numerical data. The

coding is one way to encode amino acids or nucleic acids [Wu,

95; Kawabata and Doi, 1997; White and Seffens, 1998]. The

sis kernel function is therefore inappropriate to use. The dot

kernel function can be an alternative and it is defined as below,

ߖሺܠ, ܠ௡ሻൌܠ^࢚ܠ௡

(3.72)

(a) (b) (c)

The ROC curves of a SVM model for the binary-encoded factor Xa protease

ata. (a) Use the dot product kernel. The AUC was 92.9. (b) Use the sigmoid

AUC was 92.4. (c) Use the polynomial kernel. The AUC was 94.4.