<s>
The	O
problem	O
with	O
developing	O
algorithms	O
to	O
determine	O
underpricing	O
is	O
dealing	O
with	O
noisy	B-General_Concept
,	O
complex	O
,	O
and	O
unordered	O
data	O
sets	O
.	O
</s>
<s>
To	O
resolve	O
these	O
issues	O
,	O
researchers	O
have	O
found	O
various	O
techniques	O
from	O
artificial	B-Application
intelligence	I-Application
that	O
normalizes	O
the	O
data	O
.	O
</s>
<s>
Artificial	B-Architecture
neural	I-Architecture
networks	I-Architecture
(	O
ANNs	O
)	O
resolves	O
these	O
issues	O
by	O
scanning	O
the	O
data	O
to	O
develop	O
internal	O
representations	O
of	O
the	O
relationship	O
between	O
the	O
data	O
.	O
</s>
<s>
There	O
are	O
two	O
models	O
for	O
ANNs	O
:	O
supervised	B-General_Concept
learning	I-General_Concept
and	O
unsupervised	B-General_Concept
learning	I-General_Concept
.	O
</s>
<s>
In	O
supervised	B-General_Concept
learning	I-General_Concept
models	O
,	O
there	O
are	O
tests	O
that	O
are	O
needed	O
to	O
pass	O
to	O
reduce	O
mistakes	O
.	O
</s>
<s>
test	O
output	O
does	O
not	O
match	O
test	O
input	O
,	O
the	O
algorithms	O
use	O
back	B-Algorithm
propagation	I-Algorithm
to	O
fix	O
mistakes	O
.	O
</s>
<s>
Whereas	O
in	O
unsupervised	B-General_Concept
learning	I-General_Concept
models	O
,	O
the	O
input	O
is	O
classified	O
based	O
on	O
which	O
problems	O
need	O
to	O
be	O
resolved	O
.	O
</s>
<s>
Evolutionary	B-Algorithm
programming	I-Algorithm
is	O
often	O
paired	O
with	O
other	O
algorithms	O
e.g.	O
</s>
<s>
ANN	B-Architecture
to	O
improve	O
the	O
robustness	O
,	O
reliability	O
,	O
and	O
adaptability	O
.	O
</s>
<s>
The	O
algorithm	O
deals	O
with	O
the	O
data	O
by	O
allocating	O
regions	O
for	O
noisy	B-General_Concept
data	O
.	O
</s>
<s>
The	O
scheme	O
has	O
the	O
advantage	O
of	O
isolating	O
noisy	B-General_Concept
patterns	O
which	O
reduces	O
the	O
effect	O
outliers	O
have	O
on	O
the	O
rule-generation	O
system	O
.	O
</s>
