mma and coef0 are two parameters for different kernel function

nd kernel is used to define the type of a kernel function,

el=svm(formula,data, gamma,coef0,kernel)

An illustration of support vectors using SVM for a simulated data set. The filled

e crosses stand for two classes of data. The large circles stand for support vectors.

onstruct a SVM model, there are a few things which require

They are the kernel function and the kernel parameters of the

unction. The radial basis kernel with a parameter ߛ has been

ed in applications for most numerical data, in which the distances

data points can be calculated by the Euclidean distance. The

sis kernel function is defined as below,

ߖሺܠ, ܠ௡ሻൌexpሺെߛሺܠെܠ௡ሻ^௧ሺܠെܠ௡ሻሻ

(3.69)

ast cancer data set [Wolberg, et al., 1994; Wolberg, et al., 1995]

for the demonstration. To optimise the ߛ parameter, a grid search

for ߛ to vary from 0.001 to 0.1. Figure 3.31(a) shows how the

s varying along with the ߛ values. Figure 3.31(b) shows the ROC

this SVM model employing the best ߛ value obtained by a grid

hich was 0.053.

olynomial kernel function used in SVM is defined using three

rs shown as below, where d, ߙ and ߚ are two parameters,

ߖሺܠ, ܠ௡ሻൌሺߙܠ^࢚ܠ௡൅ߚሻ^ࢊ

(3.70)