<s>
Predictive	B-General_Concept
modelling	I-General_Concept
uses	O
statistics	O
to	O
predict	O
outcomes	O
.	O
</s>
<s>
Most	O
often	O
the	O
event	O
one	O
wants	O
to	O
predict	O
is	O
in	O
the	O
future	O
,	O
but	O
predictive	B-General_Concept
modelling	I-General_Concept
can	O
be	O
applied	O
to	O
any	O
type	O
of	O
unknown	O
event	O
,	O
regardless	O
of	O
when	O
it	O
occurred	O
.	O
</s>
<s>
For	O
example	O
,	O
predictive	B-General_Concept
models	I-General_Concept
are	O
often	O
used	O
to	O
detect	O
crimes	O
and	O
identify	O
suspects	O
,	O
after	O
the	O
crime	O
has	O
taken	O
place	O
.	O
</s>
<s>
Models	O
can	O
use	O
one	O
or	O
more	O
classifiers	B-General_Concept
in	O
trying	O
to	O
determine	O
the	O
probability	O
of	O
a	O
set	O
of	O
data	O
belonging	O
to	O
another	O
set	O
.	O
</s>
<s>
Depending	O
on	O
definitional	O
boundaries	O
,	O
predictive	B-General_Concept
modelling	I-General_Concept
is	O
synonymous	O
with	O
,	O
or	O
largely	O
overlapping	O
with	O
,	O
the	O
field	O
of	O
machine	O
learning	O
,	O
as	O
it	O
is	O
more	O
commonly	O
referred	O
to	O
in	O
academic	O
or	O
research	O
and	O
development	O
contexts	O
.	O
</s>
<s>
When	O
deployed	O
commercially	O
,	O
predictive	B-General_Concept
modelling	I-General_Concept
is	O
often	O
referred	O
to	O
as	O
predictive	B-General_Concept
analytics	I-General_Concept
.	O
</s>
<s>
Predictive	B-General_Concept
modelling	I-General_Concept
is	O
often	O
contrasted	O
with	O
causal	O
modelling/analysis	O
.	O
</s>
<s>
Broadly	O
speaking	O
,	O
there	O
are	O
two	O
classes	O
of	O
predictive	B-General_Concept
models	I-General_Concept
:	O
parametric	B-General_Concept
and	O
non-parametric	B-General_Concept
.	O
</s>
<s>
Parametric	B-General_Concept
models	I-General_Concept
make	O
"	O
specific	O
assumptions	O
with	O
regard	O
to	O
one	O
or	O
more	O
of	O
the	O
population	O
parameters	O
that	O
characterize	O
the	O
underlying	O
distribution(s )	O
"	O
.	O
</s>
<s>
Non-parametric	B-General_Concept
models	I-General_Concept
"	O
typically	O
involve	O
fewer	O
assumptions	O
of	O
structure	O
and	O
distributional	O
form	O
[	O
than	O
parametric	O
models ]	O
but	O
usually	O
contain	O
strong	O
assumptions	O
about	O
independencies	O
"	O
.	O
</s>
<s>
Predictive	B-General_Concept
modelling	I-General_Concept
in	O
archaeology	O
gets	O
its	O
foundations	O
from	O
Gordon	O
Willey	O
's	O
mid-fifties	O
work	O
in	O
the	O
Virú	O
Valley	O
of	O
Peru	O
.	O
</s>
<s>
Generally	O
,	O
predictive	B-General_Concept
modelling	I-General_Concept
in	O
archaeology	O
is	O
establishing	O
statistically	O
valid	O
causal	O
or	O
covariable	O
relationships	O
between	O
natural	O
proxies	O
such	O
as	O
soil	O
types	O
,	O
elevation	O
,	O
slope	O
,	O
vegetation	O
,	O
proximity	O
to	O
water	O
,	O
geology	O
,	O
geomorphology	O
,	O
etc.	O
,	O
and	O
the	O
presence	O
of	O
archaeological	O
features	O
.	O
</s>
<s>
By	O
using	O
predictive	B-General_Concept
modelling	I-General_Concept
in	O
their	O
cultural	O
resource	O
management	O
plans	O
,	O
they	O
are	O
capable	O
of	O
making	O
more	O
informed	O
decisions	O
when	O
planning	O
for	O
activities	O
that	O
have	O
the	O
potential	O
to	O
require	O
ground	O
disturbance	O
and	O
subsequently	O
affect	O
archaeological	O
sites	O
.	O
</s>
<s>
Predictive	B-General_Concept
modelling	I-General_Concept
is	O
used	O
extensively	O
in	O
analytical	O
customer	O
relationship	O
management	O
and	O
data	B-Application
mining	I-Application
to	O
produce	O
customer-level	O
models	O
that	O
describe	O
the	O
likelihood	O
that	O
a	O
customer	O
will	O
take	O
a	O
particular	O
action	O
.	O
</s>
<s>
For	O
example	O
,	O
a	O
large	O
consumer	O
organization	O
such	O
as	O
a	O
mobile	O
telecommunications	O
operator	O
will	O
have	O
a	O
set	O
of	O
predictive	B-General_Concept
models	I-General_Concept
for	O
product	O
cross-sell	O
,	O
product	O
deep-sell	O
(	O
or	O
upselling	O
)	O
and	O
churn	O
.	O
</s>
<s>
Predictive	B-General_Concept
modelling	I-General_Concept
is	O
utilised	O
in	O
vehicle	O
insurance	O
to	O
assign	O
risk	O
of	O
incidents	O
to	O
policy	O
holders	O
from	O
information	O
obtained	O
from	O
policy	O
holders	O
.	O
</s>
<s>
This	O
is	O
extensively	O
employed	O
in	O
usage-based	O
insurance	O
solutions	O
where	O
predictive	B-General_Concept
models	I-General_Concept
utilise	O
telemetry-based	O
data	O
to	O
build	O
a	O
model	O
of	O
predictive	O
risk	O
for	O
claim	O
likelihood	O
.	O
</s>
<s>
Black-box	O
auto	O
insurance	O
predictive	B-General_Concept
models	I-General_Concept
utilise	O
GPS	O
or	O
accelerometer	O
sensor	O
input	O
only	O
.	O
</s>
<s>
In	O
2009	O
Parkland	O
Health	O
&	O
Hospital	O
System	O
began	O
analyzing	O
electronic	B-Application
medical	I-Application
records	I-Application
in	O
order	O
to	O
use	O
predictive	B-General_Concept
modeling	I-General_Concept
to	O
help	O
identify	O
patients	O
at	O
high	O
risk	O
of	O
readmission	O
.	O
</s>
<s>
proposed	O
a	O
deep	B-Algorithm
learning	I-Algorithm
model	O
for	O
estimating	O
short-term	O
life	O
expectancy	O
(	O
>3	O
months	O
)	O
of	O
the	O
patients	O
by	O
analyzing	O
free-text	O
clinical	O
notes	O
in	O
the	O
electronic	B-Application
medical	I-Application
record	I-Application
,	O
while	O
maintaining	O
the	O
temporal	O
visit	O
sequence	O
.	O
</s>
<s>
It	O
achieved	O
an	O
area	O
under	O
the	O
ROC	O
(	O
Receiver	B-Algorithm
Operating	I-Algorithm
Characteristic	I-Algorithm
)	O
curve	O
of	O
0.89	O
.	O
</s>
<s>
Predictive	B-General_Concept
modelling	I-General_Concept
has	O
been	O
used	O
to	O
estimate	O
surgery	O
duration	O
.	O
</s>
<s>
Predictive	B-General_Concept
modeling	I-General_Concept
in	O
trading	O
is	O
a	O
modeling	O
process	O
wherein	O
the	O
probability	O
of	O
an	O
outcome	O
is	O
predicted	O
using	O
a	O
set	O
of	O
predictor	O
variables	O
.	O
</s>
<s>
Predictive	B-General_Concept
models	I-General_Concept
can	O
be	O
built	O
for	O
different	O
assets	O
like	O
stocks	O
,	O
futures	O
,	O
currencies	O
,	O
commodities	O
etc	O
.	O
</s>
<s>
Predictive	B-General_Concept
modeling	I-General_Concept
is	O
still	O
extensively	O
used	O
by	O
trading	O
firms	O
to	O
devise	O
strategies	O
and	O
trade	O
.	O
</s>
<s>
Predictive	B-General_Concept
modelling	I-General_Concept
gives	O
lead	O
generators	O
a	O
head	O
start	O
by	O
forecasting	O
data-driven	O
outcomes	O
for	O
each	O
potential	O
campaign	O
.	O
</s>
<s>
Although	O
not	O
widely	O
discussed	O
by	O
the	O
mainstream	O
predictive	B-General_Concept
modeling	I-General_Concept
community	O
,	O
predictive	B-General_Concept
modeling	I-General_Concept
is	O
a	O
methodology	O
that	O
has	O
been	O
widely	O
used	O
in	O
the	O
financial	O
industry	O
in	O
the	O
past	O
and	O
some	O
of	O
the	O
major	O
failures	O
contributed	O
to	O
the	O
financial	O
crisis	O
of	O
2007	O
–	O
2008	O
.	O
</s>
<s>
The	O
models	O
produced	O
impressive	O
profits	O
until	O
a	O
major	O
debacle	O
that	O
caused	O
the	O
then	O
Federal	O
Reserve	O
chairman	O
Alan	O
Greenspan	O
to	O
step	O
in	O
to	O
broker	O
a	O
rescue	O
plan	O
by	O
the	O
Wall	B-Application
Street	I-Application
broker	O
dealers	O
in	O
order	O
to	O
prevent	O
a	O
meltdown	O
of	O
the	O
bond	O
market	O
.	O
</s>
