<s>
Data	B-General_Concept
wrangling	I-General_Concept
,	O
sometimes	O
referred	O
to	O
as	O
data	B-General_Concept
munging	I-General_Concept
,	O
is	O
the	O
process	O
of	O
transforming	O
and	O
mapping	B-General_Concept
data	I-General_Concept
from	O
one	O
"	O
raw	O
"	O
data	O
form	O
into	O
another	O
format	O
with	O
the	O
intent	O
of	O
making	O
it	O
more	O
appropriate	O
and	O
valuable	O
for	O
a	O
variety	O
of	O
downstream	O
purposes	O
such	O
as	O
analytics	O
.	O
</s>
<s>
The	O
goal	O
of	O
data	B-General_Concept
wrangling	I-General_Concept
is	O
to	O
assure	O
quality	O
and	O
useful	O
data	O
.	O
</s>
<s>
Data	O
analysts	O
typically	O
spend	O
the	O
majority	O
of	O
their	O
time	O
in	O
the	O
process	O
of	O
data	B-General_Concept
wrangling	I-General_Concept
compared	O
to	O
the	O
actual	O
analysis	O
of	O
the	O
data	O
.	O
</s>
<s>
The	O
process	O
of	O
data	B-General_Concept
wrangling	I-General_Concept
may	O
include	O
further	O
munging	O
,	O
data	B-Application
visualization	I-Application
,	O
data	O
aggregation	O
,	O
training	O
a	O
statistical	O
model	O
,	O
as	O
well	O
as	O
many	O
other	O
potential	O
uses	O
.	O
</s>
<s>
Data	B-General_Concept
wrangling	I-General_Concept
typically	O
follows	O
a	O
set	O
of	O
general	O
steps	O
which	O
begin	O
with	O
extracting	O
the	O
data	O
in	O
a	O
raw	O
form	O
from	O
the	O
data	O
source	O
,	O
"	O
munging	O
"	O
the	O
raw	O
data	O
(	O
e.g.	O
</s>
<s>
The	O
term	O
"	O
data	B-General_Concept
wrangler	I-General_Concept
"	O
was	O
also	O
suggested	O
as	O
the	O
best	O
analogy	O
to	O
describe	O
someone	O
working	O
with	O
data	O
.	O
</s>
<s>
One	O
of	O
the	O
first	O
mentions	O
of	O
data	B-General_Concept
wrangling	I-General_Concept
in	O
a	O
scientific	O
context	O
was	O
by	O
Donald	O
Cline	O
during	O
the	O
NASA/NOAA	O
Cold	O
Lands	O
Processes	O
Experiment	O
.	O
</s>
<s>
Cline	O
stated	O
the	O
data	B-General_Concept
wranglers	I-General_Concept
"	O
coordinate	O
the	O
acquisition	O
of	O
the	O
entire	O
collection	O
of	O
the	O
experiment	O
data.	O
"	O
</s>
<s>
With	O
the	O
upcoming	O
of	O
artificial	O
intelligence	O
in	O
data	O
science	O
it	O
has	O
become	O
increasingly	O
important	O
for	O
automation	O
of	O
data	B-General_Concept
wrangling	I-General_Concept
to	O
have	O
very	O
strict	O
checks	O
and	O
balances	O
,	O
which	O
is	O
why	O
the	O
munging	O
process	O
of	O
data	O
has	O
not	O
been	O
automated	O
by	O
machine	O
learning	O
.	O
</s>
<s>
Data	B-General_Concept
munging	I-General_Concept
requires	O
more	O
than	O
just	O
an	O
automated	O
solution	O
,	O
it	O
requires	O
knowledge	O
of	O
what	O
information	O
should	O
be	O
removed	O
and	O
artificial	O
intelligence	O
is	O
not	O
to	O
the	O
point	O
of	O
understanding	O
such	O
things	O
.	O
</s>
<s>
Data	B-General_Concept
wrangling	I-General_Concept
is	O
a	O
superset	O
of	O
data	B-Application
mining	I-Application
and	O
requires	O
processes	O
that	O
some	O
data	B-Application
mining	I-Application
uses	O
,	O
but	O
not	O
always	O
.	O
</s>
<s>
The	O
process	O
of	O
data	B-Application
mining	I-Application
is	O
to	O
find	O
patterns	O
within	O
large	O
data	O
sets	O
,	O
where	O
data	B-General_Concept
wrangling	I-General_Concept
transforms	O
data	O
in	O
order	O
to	O
deliver	O
insights	O
about	O
that	O
data	O
.	O
</s>
<s>
Even	O
though	O
data	B-General_Concept
wrangling	I-General_Concept
is	O
a	O
superset	O
of	O
data	B-Application
mining	I-Application
does	O
not	O
mean	O
that	O
data	B-Application
mining	I-Application
does	O
not	O
use	O
it	O
,	O
there	O
are	O
many	O
use	O
cases	O
for	O
data	B-General_Concept
wrangling	I-General_Concept
in	O
data	B-Application
mining	I-Application
.	O
</s>
<s>
Data	B-General_Concept
wrangling	I-General_Concept
can	O
benefit	O
data	B-Application
mining	I-Application
by	O
removing	O
data	O
that	O
does	O
not	O
benefit	O
the	O
overall	O
set	O
,	O
or	O
is	O
not	O
formatted	O
properly	O
,	O
which	O
will	O
yield	O
better	O
results	O
for	O
the	O
overall	O
data	B-Application
mining	I-Application
process	O
.	O
</s>
<s>
An	O
example	O
of	O
data	B-Application
mining	I-Application
that	O
is	O
closely	O
related	O
to	O
data	B-General_Concept
wrangling	I-General_Concept
is	O
ignoring	O
data	O
from	O
a	O
set	O
that	O
is	O
not	O
connected	O
to	O
the	O
goal	O
:	O
say	O
there	O
is	O
a	O
data	O
set	O
related	O
to	O
the	O
state	O
of	O
Texas	O
and	O
the	O
goal	O
is	O
to	O
get	O
statistics	O
on	O
the	O
residents	O
of	O
Houston	O
,	O
the	O
data	O
in	O
the	O
set	O
related	O
to	O
the	O
residents	O
of	O
Dallas	O
is	O
not	O
useful	O
to	O
the	O
overall	O
set	O
and	O
can	O
be	O
removed	O
before	O
processing	O
to	O
improve	O
the	O
efficiency	O
of	O
the	O
data	B-Application
mining	I-Application
process	O
.	O
</s>
<s>
With	O
an	O
increase	O
of	O
raw	O
data	O
comes	O
an	O
increase	O
in	O
the	O
amount	O
of	O
data	O
that	O
is	O
not	O
inherently	O
useful	O
,	O
this	O
increases	O
time	O
spent	O
on	O
cleaning	O
and	O
organizing	O
data	O
before	O
it	O
can	O
be	O
analyzed	O
which	O
is	O
where	O
data	B-General_Concept
wrangling	I-General_Concept
comes	O
into	O
play	O
.	O
</s>
<s>
The	O
result	O
of	O
data	B-General_Concept
wrangling	I-General_Concept
can	O
provide	O
important	O
metadata	O
statistics	O
for	O
further	O
insights	O
about	O
the	O
data	O
,	O
it	O
is	O
important	O
to	O
ensure	O
metadata	O
is	O
consistent	O
otherwise	O
it	O
can	O
cause	O
roadblocks	O
.	O
</s>
<s>
Data	B-General_Concept
wrangling	I-General_Concept
allows	O
analysts	O
to	O
analyze	O
more	O
complex	O
data	O
more	O
quickly	O
,	O
achieve	O
more	O
accurate	O
results	O
,	O
and	O
because	O
of	O
this	O
better	O
decisions	O
can	O
be	O
made	O
.	O
</s>
<s>
Many	O
businesses	O
have	O
moved	O
to	O
data	B-General_Concept
wrangling	I-General_Concept
because	O
of	O
the	O
success	O
that	O
it	O
has	O
brought	O
.	O
</s>
<s>
The	O
main	O
steps	O
in	O
data	B-General_Concept
wrangling	I-General_Concept
are	O
as	O
follows	O
:	O
</s>
<s>
The	O
result	O
of	O
using	O
the	O
data	B-General_Concept
wrangling	I-General_Concept
process	O
on	O
this	O
small	O
data	O
set	O
shows	O
a	O
significantly	O
easier	O
data	O
set	O
to	O
read	O
.	O
</s>
<s>
The	O
recipients	O
could	O
be	O
individuals	O
,	O
such	O
as	O
data	B-Application
architects	I-Application
or	O
data	O
scientists	O
who	O
will	O
investigate	O
the	O
data	O
further	O
,	O
business	O
users	O
who	O
will	O
consume	O
the	O
data	O
directly	O
in	O
reports	O
,	O
or	O
systems	O
that	O
will	O
further	O
process	O
the	O
data	O
and	O
write	O
it	O
into	O
targets	O
such	O
as	O
data	O
warehouses	O
,	O
data	B-General_Concept
lakes	I-General_Concept
,	O
or	O
downstream	O
applications	O
.	O
</s>
<s>
Depending	O
on	O
the	O
amount	O
and	O
format	O
of	O
the	O
incoming	O
data	O
,	O
data	B-General_Concept
wrangling	I-General_Concept
has	O
traditionally	O
been	O
performed	O
manually	O
(	O
e.g.	O
</s>
<s>
via	O
spreadsheets	O
such	O
as	O
Excel	O
)	O
,	O
tools	O
like	O
KNIME	B-Language
or	O
via	O
scripts	O
in	O
languages	O
such	O
as	O
Python	B-Language
or	O
SQL	B-Language
.	O
</s>
<s>
R	B-Language
,	O
a	O
language	O
often	O
used	O
in	O
data	B-Application
mining	I-Application
and	O
statistical	O
data	O
analysis	O
,	O
is	O
now	O
also	O
sometimes	O
used	O
for	O
data	B-General_Concept
wrangling	I-General_Concept
.	O
</s>
<s>
Data	B-General_Concept
wranglers	I-General_Concept
typically	O
have	O
skills	O
sets	O
within	O
:	O
R	B-Language
or	O
Python	B-Language
,	O
SQL	B-Language
,	O
PHP	O
,	O
Scala	O
,	O
and	O
more	O
languages	O
typically	O
used	O
for	O
analyzing	O
data	O
.	O
</s>
<s>
Visual	O
data	B-General_Concept
wrangling	I-General_Concept
systems	O
were	O
developed	O
to	O
make	O
data	B-General_Concept
wrangling	I-General_Concept
accessible	O
for	O
non-programmers	O
,	O
and	O
simpler	O
for	O
programmers	O
.	O
</s>
<s>
Some	O
of	O
these	O
also	O
include	O
embedded	O
AI	O
recommenders	B-Application
and	O
programming	B-Application
by	I-Application
example	I-Application
facilities	O
to	O
provide	O
user	O
assistance	O
,	O
and	O
program	B-Application
synthesis	I-Application
techniques	O
to	O
autogenerate	O
scalable	O
dataflow	O
code	O
.	O
</s>
<s>
Early	O
prototypes	O
of	O
visual	O
data	B-General_Concept
wrangling	I-General_Concept
tools	O
include	O
OpenRefine	B-Application
and	O
the	O
Stanford/Berkeley	O
research	O
system	O
;	O
the	O
latter	O
evolved	O
into	O
Trifacta	O
.	O
</s>
<s>
Other	O
terms	O
for	O
these	O
processes	O
have	O
included	O
data	O
franchising	O
,	O
data	B-General_Concept
preparation	I-General_Concept
,	O
and	O
data	B-General_Concept
munging	I-General_Concept
.	O
</s>
<s>
Once	O
an	O
understanding	O
of	O
the	O
outcome	O
is	O
achieved	O
then	O
the	O
data	B-General_Concept
wrangling	I-General_Concept
process	O
can	O
begin	O
.	O
</s>
