|
In this
example, a
new credit
scoring
system is
applied to
old
historical
data from a
check
guarantee
company.
Bad checks
are
distributed
at the low
end of the
score
spectrum,
while good
checks are
distributed
at the high
end.
Graph1
shows
separation
between
goods
and bads
with any
credit
scoring
system.
Graph2
shows
increased
separation
with
neural
networks.
Graph3
shows
added
profit
potential
from
better
separation.
|
