<s>
"	O
A	O
"	O
,	O
"	O
B	O
"	O
,	O
"	O
AB	O
"	O
or	O
"	O
O	O
"	O
,	O
for	O
blood	O
type	O
)	O
,	O
ordinal	B-General_Concept
(	O
e.g.	O
</s>
<s>
Other	O
classifiers	B-General_Concept
work	O
by	O
comparing	O
observations	O
to	O
previous	O
observations	O
by	O
means	O
of	O
a	O
similarity	O
or	O
distance	O
function	O
.	O
</s>
<s>
An	O
algorithm	O
that	O
implements	O
classification	O
,	O
especially	O
in	O
a	O
concrete	O
implementation	O
,	O
is	O
known	O
as	O
a	O
classifier	B-General_Concept
.	O
</s>
<s>
The	O
term	O
"	O
classifier	B-General_Concept
"	O
sometimes	O
also	O
refers	O
to	O
the	O
mathematical	O
function	O
,	O
implemented	O
by	O
a	O
classification	O
algorithm	O
,	O
that	O
maps	O
input	O
data	O
to	O
a	O
category	O
.	O
</s>
<s>
In	O
machine	O
learning	O
,	O
the	O
observations	O
are	O
often	O
known	O
as	O
instances	O
,	O
the	O
explanatory	O
variables	O
are	O
termed	O
features	O
(	O
grouped	O
into	O
a	O
feature	B-Algorithm
vector	I-Algorithm
)	O
,	O
and	O
the	O
possible	O
categories	O
to	O
be	O
predicted	O
are	O
classes	O
.	O
</s>
<s>
in	O
community	O
ecology	O
,	O
the	O
term	O
"	O
classification	O
"	O
normally	O
refers	O
to	O
cluster	B-Algorithm
analysis	I-Algorithm
.	O
</s>
<s>
Other	O
examples	O
are	O
regression	O
,	O
which	O
assigns	O
a	O
real-valued	O
output	O
to	O
each	O
input	O
;	O
sequence	B-General_Concept
labeling	I-General_Concept
,	O
which	O
assigns	O
a	O
class	O
to	O
each	O
member	O
of	O
a	O
sequence	O
of	O
values	O
(	O
for	O
example	O
,	O
part	O
of	O
speech	O
tagging	O
,	O
which	O
assigns	O
a	O
part	O
of	O
speech	O
to	O
each	O
word	O
in	O
an	O
input	O
sentence	O
)	O
;	O
parsing	B-Language
,	O
which	O
assigns	O
a	O
parse	B-Language
tree	O
to	O
an	O
input	O
sentence	O
,	O
describing	O
the	O
syntactic	O
structure	O
of	O
the	O
sentence	O
;	O
etc	O
.	O
</s>
<s>
A	O
common	O
subclass	O
of	O
classification	O
is	O
probabilistic	B-General_Concept
classification	I-General_Concept
.	O
</s>
<s>
It	O
can	O
output	O
a	O
confidence	O
value	O
associated	O
with	O
its	O
choice	O
(	O
in	O
general	O
,	O
a	O
classifier	B-General_Concept
that	O
can	O
do	O
this	O
is	O
known	O
as	O
a	O
confidence-weighted	O
classifier	B-General_Concept
)	O
.	O
</s>
<s>
Because	O
of	O
the	O
probabilities	O
which	O
are	O
generated	O
,	O
probabilistic	B-General_Concept
classifiers	I-General_Concept
can	O
be	O
more	O
effectively	O
incorporated	O
into	O
larger	O
machine-learning	O
tasks	O
,	O
in	O
a	O
way	O
that	O
partially	O
or	O
completely	O
avoids	O
the	O
problem	O
of	O
error	O
propagation	O
.	O
</s>
<s>
Early	O
work	O
on	O
statistical	B-General_Concept
classification	I-General_Concept
was	O
undertaken	O
by	O
Fisher	O
,	O
in	O
the	O
context	O
of	O
two-group	O
problems	O
,	O
leading	O
to	O
Fisher	B-General_Concept
's	I-General_Concept
linear	I-General_Concept
discriminant	I-General_Concept
function	O
as	O
the	O
rule	O
for	O
assigning	O
a	O
group	O
to	O
a	O
new	O
observation	O
.	O
</s>
<s>
Later	O
work	O
for	O
the	O
multivariate	O
normal	O
distribution	O
allowed	O
the	O
classifier	B-General_Concept
to	O
be	O
nonlinear	O
:	O
several	O
classification	O
rules	O
can	O
be	O
derived	O
based	O
on	O
different	O
adjustments	O
of	O
the	O
Mahalanobis	O
distance	O
,	O
with	O
a	O
new	O
observation	O
being	O
assigned	O
to	O
the	O
group	O
whose	O
centre	O
has	O
the	O
lowest	O
adjusted	O
distance	O
from	O
the	O
observation	O
.	O
</s>
<s>
Bayesian	O
procedures	O
tend	O
to	O
be	O
computationally	O
expensive	O
and	O
,	O
in	O
the	O
days	O
before	O
Markov	B-General_Concept
chain	I-General_Concept
Monte	I-General_Concept
Carlo	I-General_Concept
computations	O
were	O
developed	O
,	O
approximations	O
for	O
Bayesian	O
clustering	O
rules	O
were	O
devised	O
.	O
</s>
<s>
Some	O
Bayesian	O
procedures	O
involve	O
the	O
calculation	O
of	O
group	B-General_Concept
membership	I-General_Concept
probabilities	I-General_Concept
:	O
these	O
provide	O
a	O
more	O
informative	O
outcome	O
than	O
a	O
simple	O
attribution	O
of	O
a	O
single	O
group-label	O
to	O
each	O
new	O
observation	O
.	O
</s>
<s>
Classification	O
can	O
be	O
thought	O
of	O
as	O
two	O
separate	O
problems	O
–	O
binary	B-General_Concept
classification	I-General_Concept
and	O
multiclass	B-General_Concept
classification	I-General_Concept
.	O
</s>
<s>
In	O
binary	B-General_Concept
classification	I-General_Concept
,	O
a	O
better	O
understood	O
task	O
,	O
only	O
two	O
classes	O
are	O
involved	O
,	O
whereas	O
multiclass	B-General_Concept
classification	I-General_Concept
involves	O
assigning	O
an	O
object	O
to	O
one	O
of	O
several	O
classes	O
.	O
</s>
<s>
Since	O
many	O
classification	O
methods	O
have	O
been	O
developed	O
specifically	O
for	O
binary	B-General_Concept
classification	I-General_Concept
,	O
multiclass	B-General_Concept
classification	I-General_Concept
often	O
requires	O
the	O
combined	O
use	O
of	O
multiple	O
binary	B-General_Concept
classifiers	I-General_Concept
.	O
</s>
<s>
Most	O
algorithms	O
describe	O
an	O
individual	O
instance	O
whose	O
category	O
is	O
to	O
be	O
predicted	O
using	O
a	O
feature	B-Algorithm
vector	I-Algorithm
of	O
individual	O
,	O
measurable	O
properties	O
of	O
the	O
instance	O
.	O
</s>
<s>
Each	O
property	O
is	O
termed	O
a	O
feature	B-Algorithm
,	O
also	O
known	O
in	O
statistics	O
as	O
an	O
explanatory	O
variable	O
(	O
or	O
independent	O
variable	O
,	O
although	O
features	O
may	O
or	O
may	O
not	O
be	O
statistically	O
independent	O
)	O
.	O
</s>
<s>
"	O
A	O
"	O
,	O
"	O
B	O
"	O
,	O
"	O
AB	O
"	O
or	O
"	O
O	O
"	O
,	O
for	O
blood	O
type	O
)	O
;	O
ordinal	B-General_Concept
(	O
e.g.	O
</s>
<s>
If	O
the	O
instance	O
is	O
an	O
image	O
,	O
the	O
feature	B-Algorithm
values	O
might	O
correspond	O
to	O
the	O
pixels	O
of	O
an	O
image	O
;	O
if	O
the	O
instance	O
is	O
a	O
piece	O
of	O
text	O
,	O
the	O
feature	B-Algorithm
values	O
might	O
be	O
occurrence	O
frequencies	O
of	O
different	O
words	O
.	O
</s>
<s>
A	O
large	O
number	O
of	O
algorithms	O
for	O
classification	O
can	O
be	O
phrased	O
in	O
terms	O
of	O
a	O
linear	O
function	O
that	O
assigns	O
a	O
score	O
to	O
each	O
possible	O
category	O
k	O
by	O
combining	O
the	O
feature	B-Algorithm
vector	I-Algorithm
of	O
an	O
instance	O
with	O
a	O
vector	O
of	O
weights	O
,	O
using	O
a	O
dot	O
product	O
.	O
</s>
<s>
This	O
type	O
of	O
score	O
function	O
is	O
known	O
as	O
a	O
linear	B-General_Concept
predictor	I-General_Concept
function	I-General_Concept
and	O
has	O
the	O
following	O
general	O
form	O
:	O
</s>
<s>
where	O
Xi	O
is	O
the	O
feature	B-Algorithm
vector	I-Algorithm
for	O
instance	O
i	O
,	O
βk	O
is	O
the	O
vector	O
of	O
weights	O
corresponding	O
to	O
category	O
k	O
,	O
and	O
score(Xi, k )	O
is	O
the	O
score	O
associated	O
with	O
assigning	O
instance	O
i	O
to	O
category	O
k	O
.	O
In	O
discrete	O
choice	O
theory	O
,	O
where	O
instances	O
represent	O
people	O
and	O
categories	O
represent	O
choices	O
,	O
the	O
score	O
is	O
considered	O
the	O
utility	O
associated	O
with	O
person	O
i	O
choosing	O
category	O
k	O
.	O
</s>
<s>
Algorithms	O
with	O
this	O
basic	O
setup	O
are	O
known	O
as	O
linear	B-General_Concept
classifiers	I-General_Concept
.	O
</s>
<s>
Classifier	B-General_Concept
performance	O
depends	O
greatly	O
on	O
the	O
characteristics	O
of	O
the	O
data	O
to	O
be	O
classified	O
.	O
</s>
<s>
There	O
is	O
no	O
single	O
classifier	B-General_Concept
that	O
works	O
best	O
on	O
all	O
given	O
problems	O
(	O
a	O
phenomenon	O
that	O
may	O
be	O
explained	O
by	O
the	O
no-free-lunch	O
theorem	O
)	O
.	O
</s>
<s>
Various	O
empirical	O
tests	O
have	O
been	O
performed	O
to	O
compare	O
classifier	B-General_Concept
performance	O
and	O
to	O
find	O
the	O
characteristics	O
of	O
data	O
that	O
determine	O
classifier	B-General_Concept
performance	O
.	O
</s>
<s>
Determining	O
a	O
suitable	O
classifier	B-General_Concept
for	O
a	O
given	O
problem	O
is	O
however	O
still	O
more	O
an	O
art	O
than	O
a	O
science	O
.	O
</s>
<s>
More	O
recently	O
,	O
receiver	B-Algorithm
operating	I-Algorithm
characteristic	I-Algorithm
(	O
ROC	O
)	O
curves	O
have	O
been	O
used	O
to	O
evaluate	O
the	O
tradeoff	O
between	O
true	O
-	O
and	O
false-positive	O
rates	O
of	O
classification	O
algorithms	O
.	O
</s>
<s>
In	O
some	O
of	O
these	O
it	O
is	O
employed	O
as	O
a	O
data	B-Application
mining	I-Application
procedure	O
,	O
while	O
in	O
others	O
more	O
detailed	O
statistical	O
modeling	O
is	O
undertaken	O
.	O
</s>
