<s>
Classifier	B-Algorithm
chains	I-Algorithm
is	O
a	O
machine	O
learning	O
method	O
for	O
problem	O
transformation	O
in	O
multi-label	B-Algorithm
classification	I-Algorithm
.	O
</s>
<s>
It	O
combines	O
the	O
computational	O
efficiency	O
of	O
the	O
Binary	O
Relevance	O
method	O
while	O
still	O
being	O
able	O
to	O
take	O
the	O
label	O
dependencies	O
into	O
account	O
for	O
classification	B-General_Concept
.	O
</s>
<s>
Given	O
a	O
set	O
of	O
labels	O
and	O
a	O
data	O
set	O
with	O
instances	O
of	O
the	O
form	O
where	O
is	O
a	O
feature	B-Algorithm
vector	I-Algorithm
and	O
is	O
a	O
set	O
of	O
labels	O
assigned	O
to	O
the	O
instance	O
.	O
</s>
<s>
BR	O
transforms	O
the	O
data	O
set	O
into	O
data	O
sets	O
and	O
learns	O
binary	O
classifiers	B-General_Concept
for	O
each	O
label	O
.	O
</s>
<s>
Loss	O
of	O
this	O
information	O
can	O
in	O
some	O
cases	O
lead	O
to	O
a	O
decrease	O
in	O
classification	B-General_Concept
performance	O
.	O
</s>
<s>
After	O
transformation	O
a	O
single-label	O
classifier	B-General_Concept
is	O
trained	O
where	O
is	O
the	O
power	O
set	O
of	O
all	O
labels	O
in	O
.	O
</s>
<s>
For	O
example	O
,	O
a	O
multi-label	B-Algorithm
data	O
set	O
with	O
10	O
labels	O
can	O
have	O
up	O
to	O
label	O
combinations	O
.	O
</s>
<s>
This	O
increases	O
the	O
run-time	O
of	O
classification	B-General_Concept
.	O
</s>
<s>
The	O
Classifier	B-Algorithm
Chains	I-Algorithm
method	O
is	O
based	O
on	O
the	O
BR	O
method	O
and	O
it	O
is	O
efficient	O
even	O
on	O
a	O
big	O
number	O
of	O
labels	O
.	O
</s>
<s>
For	O
a	O
given	O
a	O
set	O
of	O
labels	O
the	O
Classifier	B-General_Concept
Chain	O
model	O
(	O
CC	O
)	O
learns	O
classifiers	B-General_Concept
as	O
in	O
the	O
Binary	O
Relevance	O
method	O
.	O
</s>
<s>
All	O
classifiers	B-General_Concept
are	O
linked	O
in	O
a	O
chain	O
through	O
feature	O
space	O
.	O
</s>
<s>
Thus	O
,	O
classifiers	B-General_Concept
build	O
a	O
chain	O
where	O
each	O
of	O
them	O
learns	O
binary	O
classification	B-General_Concept
of	O
a	O
single	O
label	O
.	O
</s>
<s>
The	O
features	O
given	O
to	O
each	O
classifier	B-General_Concept
are	O
extended	O
with	O
binary	O
values	O
that	O
indicate	O
which	O
of	O
previous	O
labels	O
were	O
assigned	O
to	O
the	O
instance	O
.	O
</s>
<s>
By	O
classifying	O
new	O
instances	O
the	O
labels	O
are	O
again	O
predicted	O
by	O
building	O
a	O
chain	O
of	O
classifiers	B-General_Concept
.	O
</s>
<s>
The	O
classification	B-General_Concept
begins	O
with	O
the	O
first	O
classifier	B-General_Concept
and	O
proceeds	O
to	O
the	O
last	O
one	O
by	O
passing	O
label	O
information	O
between	O
classifiers	B-General_Concept
through	O
the	O
feature	O
space	O
.	O
</s>
<s>
For	O
example	O
,	O
if	O
a	O
label	O
often	O
co-occur	O
with	O
some	O
other	O
label	O
,	O
then	O
only	O
instances	O
of	O
the	O
label	O
which	O
comes	O
later	O
in	O
the	O
chain	O
will	O
have	O
information	O
about	O
the	O
other	O
one	O
in	O
its	O
feature	B-Algorithm
vector	I-Algorithm
.	O
</s>
<s>
In	O
order	O
to	O
solve	O
this	O
problem	O
and	O
increase	O
accuracy	O
it	O
is	O
possible	O
to	O
use	O
ensemble	B-Algorithm
of	O
classifiers	B-General_Concept
.	O
</s>
<s>
In	O
Ensemble	B-Algorithm
of	O
Classifier	B-Algorithm
Chains	I-Algorithm
(	O
ECC	O
)	O
several	O
CC	O
classifiers	B-General_Concept
can	O
be	O
trained	O
with	O
random	O
order	O
of	O
chains	O
(	O
i.e.	O
</s>
<s>
Labels	O
of	O
a	O
new	O
instance	O
are	O
predicted	O
by	O
each	O
classifier	B-General_Concept
separately	O
.	O
</s>
<s>
The	O
label	O
is	O
accepted	O
if	O
it	O
was	O
predicted	O
by	O
a	O
percentage	O
of	O
classifiers	B-General_Concept
that	O
is	O
bigger	O
than	O
some	O
threshold	O
value	O
.	O
</s>
