<s>
It	O
works	O
based	O
on	O
the	O
concept	O
of	O
separate-and-conquer	O
to	O
directly	O
induce	B-General_Concept
rules	I-General_Concept
from	O
a	O
given	O
training	O
set	O
and	O
build	O
its	O
knowledge	O
repository	O
.	O
</s>
<s>
Algorithms	O
under	O
RULES	O
family	O
are	O
usually	O
available	O
in	O
data	O
mining	O
tools	O
,	O
such	O
as	O
KEEL	O
and	O
WEKA	B-Language
,	O
known	O
for	O
knowledge	O
extraction	O
and	O
decision	O
making	O
.	O
</s>
<s>
Inductive	O
learning	O
had	O
been	O
divided	O
into	O
two	O
types	O
:	O
decision	B-Algorithm
tree	I-Algorithm
(	O
DT	O
)	O
and	O
covering	O
algorithms	O
(	O
CA	O
)	O
.	O
</s>
<s>
DTs	O
discover	O
rules	O
using	O
decision	B-Algorithm
tree	I-Algorithm
based	O
on	O
the	O
concept	O
of	O
divide-and-conquer	O
,	O
while	O
CA	O
directly	O
induces	O
rules	O
from	O
the	O
training	O
set	O
based	O
on	O
the	O
concept	O
of	O
separate	O
and	O
conquers	O
.	O
</s>
<s>
Although	O
DT	O
algorithms	O
was	O
well	O
recognized	O
in	O
the	O
past	O
few	O
decades	O
,	O
CA	O
started	O
to	O
attract	O
the	O
attention	O
due	O
to	O
its	O
direct	O
rule	B-General_Concept
induction	I-General_Concept
property	O
,	O
as	O
emphasized	O
by	O
Kurgan	O
et	O
al	O
.	O
</s>
<s>
RULES	O
family	O
[2],	O
known	O
as	O
rule	B-General_Concept
extraction	I-General_Concept
system	O
,	O
is	O
one	O
family	O
of	O
covering	O
algorithms	O
that	O
separate	O
each	O
instance	O
or	O
example	O
when	O
inducing	O
the	O
best	O
rules	O
.	O
</s>
<s>
It	O
allows	O
the	O
best	O
rule	O
to	O
cover	O
some	O
negative	O
examples	O
to	O
handle	O
the	O
increase	O
flexibility	O
and	O
reduce	O
the	O
overfitting	O
problem	O
and	O
noisy	O
data	O
in	O
the	O
rule	B-General_Concept
induction	I-General_Concept
.	O
</s>
