Graphics Reference
In-Depth Information
Figure
.
.
he power curves for a perfect model, a random model, and some real classification
models. AR is the ratio of the two areas (A
B). It lies between
(a random model with no predictive
power) and
(a perfect model)
Figure
.
.
he relationship between an accuracy measure (AR) and the coe
cient r in the SVM
formulation. Higher r values correspond to less complex models. he median ARs were estimated for
training and validation sets of
bootstrapped subsamples each of
solvent and
insolvent
companies. A bivariate SVM was used with the variables K
and K
. We will be using r
=
inall
SVMs discussed in this chapter
