| Library: | xclust |
| See also: | svmplot adaptive divisive cartsplit cartcv |
| Quantlet: | svm | |
| Description: | returns the vector of scores for the objects represented in AC. AT is a training set where the last column describes the class of an object (must be +1 or -1). |
| Usage: | V=svm(AT,AC,r,c) | |
| Input: | ||
| AT | (m x d+1) matrix, contains d characteristics of m objects of the training set. The last column of the matrix must describe the class of an object, it can only be +1 or -1. Pararmeter AC | |
| r | (mm x d) matrix, contains d characteristics of mm objects whose classes (+1 or -1) are to be found. | |
| c | scalar, the coefficient determining the anisotropic radial basis r*Sigma^(1/2), where Sigma is the covariance matrix of the training set. | |
| n/a | scalar, capacity of the support vector machine invariant of the number of objects m. The capacity appearing in the SVM Lagrange formulation is C=c*m. | |
| Output: | ||
| V | (mm x 1) vector, scores. | |
library("xplore");
library("plot");
library("xclust");
AT=read("bankruptcy.dat");
nsteps=100;
origin=#(-0.62, 0.0);
endpoints=#(0.52, 1.5);
steps=(endpoints-origin)/nsteps;
npoints=#(nsteps+1, nsteps+1);
AC=grid(origin,steps,npoints);
V=svm(AT, AC, 2.0, 1.0);
; Try out these values of r and c: ;
; -------------------------------- ;
; r=100 ; c=1 ;
; r=2 ; c=1 ;
; r=0.5 ; c=1 ;
; ------------- ;
; r=2 ; c=300 ;
svmplot(AT,AC,V);
setgopt(svmdi,1,1,"xlabel","Profitability(NI/TA)","ylabel","Leverage(TL/TA)","title","Company Scores");
A two dimensional plot. Intensity of the background colour corresponds to the empirically estimated probability of default: the darker the area, the higher the company default probabilty.