Hi All,
I'm developing a ksvm application. I need to use the results obtained from
ksvm in R to construct or restore the ksvm formula and further use this formula
to predict new data.
The ksvm class contains xmatrix, ymatrix, coefficients, alpha, b and fitted
value. Taking linear svm as an example ( vanilladot), the formula is like :
F(x) = sign(sum(w*x*y*z)+b)
Where, x = xmatrix, y = ymatrix, b is intercept and z is the input matrix? The
w is alpha or coefficient from ksvm class? I did some calculation but the result
is not the same as the fitted value.
Can you give me some idea on the prediction of ksvm by the formula?
Thanks a lot!!!
Regards,
Yan
[[alternative HTML version deleted]]