<s>
In	O
network	O
theory	O
,	O
link	B-Application
analysis	I-Application
is	O
a	O
data-analysis	B-General_Concept
technique	O
used	O
to	O
evaluate	O
relationships	O
(	O
Tap	O
link	O
)	O
between	O
nodes	O
.	O
</s>
<s>
Link	B-Application
analysis	I-Application
has	O
been	O
used	O
for	O
investigation	O
of	O
criminal	O
activity	O
(	O
fraud	O
,	O
counterterrorism	O
,	O
and	O
intelligence	O
)	O
,	O
computer	O
security	O
analysis	B-General_Concept
,	O
search	O
engine	O
optimization	O
,	O
market	O
research	O
,	O
medical	O
research	O
,	O
and	O
art	O
.	O
</s>
<s>
Knowledge	O
discovery	O
is	O
an	O
iterative	B-Algorithm
and	O
interactive	B-General_Concept
process	O
used	O
to	O
identify	O
,	O
analyze	O
and	O
visualize	O
patterns	O
in	O
data	O
.	O
</s>
<s>
Network	O
analysis	B-General_Concept
,	O
link	B-Application
analysis	I-Application
and	O
social	O
network	O
analysis	B-General_Concept
are	O
all	O
methods	O
of	O
knowledge	O
discovery	O
,	O
each	O
a	O
corresponding	O
subset	O
of	O
the	O
prior	O
method	O
.	O
</s>
<s>
Manual	O
or	O
computer-generated	O
visualizations	B-Application
tools	O
may	O
be	O
mapped	O
from	O
the	O
data	O
,	O
including	O
network	O
charts	O
.	O
</s>
<s>
Several	O
algorithms	O
exist	O
to	O
help	O
with	O
analysis	B-General_Concept
of	O
data	O
–	O
Dijkstra	B-Algorithm
’s	I-Algorithm
algorithm	I-Algorithm
,	O
breadth-first	B-Algorithm
search	I-Algorithm
,	O
and	O
depth-first	B-Algorithm
search	I-Algorithm
.	O
</s>
<s>
Link	B-Application
analysis	I-Application
focuses	O
on	O
analysis	B-General_Concept
of	O
relationships	O
among	O
nodes	O
through	O
visualization	B-Application
methods	I-Application
(	O
network	O
charts	O
,	O
association	O
matrix	O
)	O
.	O
</s>
<s>
Link	B-Application
analysis	I-Application
is	O
used	O
for	O
3	O
primary	O
purposes	O
:	O
</s>
<s>
Discover	O
new	O
patterns	O
of	O
interest	O
(	O
social	O
network	O
analysis	B-General_Concept
,	O
data	B-Application
mining	I-Application
)	O
.	O
</s>
<s>
Klerks	O
categorized	O
link	B-Application
analysis	I-Application
tools	O
into	O
3	O
generations	O
.	O
</s>
<s>
This	O
method	O
requires	O
that	O
a	O
domain	O
expert	O
review	O
data	O
files	O
,	O
identify	O
associations	O
by	O
constructing	O
an	O
association	O
matrix	O
,	O
create	O
a	O
link	O
chart	O
for	O
visualization	B-Application
and	O
finally	O
analyze	O
the	O
network	O
chart	O
to	O
identify	O
patterns	O
of	O
interest	O
.	O
</s>
<s>
The	O
distinction	O
between	O
these	O
two	O
types	O
of	O
matrices	O
,	O
while	O
minor	O
,	O
is	O
nonetheless	O
significant	O
in	O
terms	O
of	O
the	O
output	O
of	O
the	O
analysis	B-General_Concept
completed	O
or	O
rendered	O
.	O
</s>
<s>
Second	O
generation	O
tools	O
consist	O
of	O
automatic	O
graphics-based	O
analysis	B-General_Concept
tools	O
such	O
as	O
IBM	O
i2	O
Analyst	O
’s	O
Notebook	O
,	O
Netmap	O
,	O
ClueMaker	O
and	O
Watson	O
.	O
</s>
<s>
These	O
tools	O
offer	O
the	O
ability	O
to	O
automate	O
the	O
construction	O
and	O
updates	O
of	O
the	O
link	O
chart	O
once	O
an	O
association	O
matrix	O
is	O
manually	O
created	O
,	O
however	O
,	O
analysis	B-General_Concept
of	O
the	O
resulting	O
charts	O
and	O
graphs	O
still	O
requires	O
an	O
expert	O
with	O
extensive	O
domain	O
knowledge	O
.	O
</s>
<s>
The	O
third	O
generation	O
of	O
link-analysis	O
tools	O
like	O
DataWalk	O
allow	O
the	O
automatic	O
visualization	B-Application
of	O
linkages	O
between	O
elements	O
in	O
a	O
data	O
set	O
,	O
that	O
can	O
then	O
serve	O
as	O
the	O
canvas	O
for	O
further	O
exploration	O
or	O
manual	O
updates	O
.	O
</s>
<s>
With	O
the	O
vast	O
amounts	O
of	O
data	O
and	O
information	O
that	O
are	O
stored	O
electronically	O
,	O
users	O
are	O
confronted	O
with	O
multiple	O
unrelated	O
sources	O
of	O
information	O
available	O
for	O
analysis	B-General_Concept
.	O
</s>
<s>
Data	B-General_Concept
analysis	I-General_Concept
techniques	O
are	O
required	O
to	O
make	O
effective	O
and	O
efficient	O
use	O
of	O
the	O
data	O
.	O
</s>
<s>
Palshikar	O
classifies	O
data	B-General_Concept
analysis	I-General_Concept
techniques	O
into	O
two	O
categories	O
–	O
(	O
statistical	O
models	O
,	O
time-series	O
analysis	B-General_Concept
,	O
clustering	O
and	O
classification	B-General_Concept
,	O
matching	O
algorithms	O
to	O
detect	O
anomalies	O
)	O
and	O
artificial	B-Application
intelligence	I-Application
(	O
AI	B-Application
)	O
techniques	O
(	O
data	B-Application
mining	I-Application
,	O
expert	B-Application
systems	I-Application
,	O
pattern	O
recognition	O
,	O
machine	O
learning	O
techniques	O
,	O
neural	B-Architecture
networks	I-Architecture
)	O
.	O
</s>
<s>
Bolton	O
&	O
Hand	O
define	O
statistical	O
data	B-General_Concept
analysis	I-General_Concept
as	O
either	O
supervised	O
or	O
unsupervised	O
methods	O
.	O
</s>
<s>
Supervised	B-General_Concept
learning	I-General_Concept
methods	I-General_Concept
require	O
that	O
rules	O
are	O
defined	O
within	O
the	O
system	O
to	O
establish	O
what	O
is	O
expected	O
or	O
unexpected	O
behavior	O
.	O
</s>
<s>
Unsupervised	B-General_Concept
learning	I-General_Concept
methods	I-General_Concept
review	O
data	O
in	O
comparison	O
to	O
the	O
norm	O
and	O
detect	O
statistical	O
outliers	O
.	O
</s>
<s>
Supervised	B-General_Concept
learning	I-General_Concept
methods	I-General_Concept
are	O
limited	O
in	O
the	O
scenarios	O
that	O
can	O
be	O
handled	O
as	O
this	O
method	O
requires	O
that	O
training	O
rules	O
are	O
established	O
based	O
on	O
previous	O
patterns	O
.	O
</s>
<s>
Unsupervised	B-General_Concept
learning	I-General_Concept
methods	I-General_Concept
can	O
provide	O
detection	O
of	O
broader	O
issues	O
,	O
however	O
,	O
may	O
result	O
in	O
a	O
higher	O
false-positive	O
ratio	O
if	O
the	O
behavioral	O
norm	O
is	O
not	O
well	O
established	O
or	O
understood	O
.	O
</s>
<s>
Sparrow	O
highlights	O
incompleteness	O
(	O
inevitability	O
of	O
missing	O
data	O
or	O
links	O
)	O
,	O
fuzzy	O
boundaries	O
(	O
subjectivity	O
in	O
deciding	O
what	O
to	O
include	O
)	O
and	O
dynamic	O
changes	O
(	O
recognition	O
that	O
data	O
is	O
ever-changing	O
)	O
as	O
the	O
three	O
primary	O
problems	O
with	O
data	B-General_Concept
analysis	I-General_Concept
.	O
</s>
<s>
The	O
primary	O
method	O
for	O
resolving	O
data	B-General_Concept
analysis	I-General_Concept
issues	O
is	O
reliance	O
on	O
domain	O
knowledge	O
from	O
an	O
expert	O
.	O
</s>
<s>
This	O
is	O
a	O
very	O
time-consuming	O
and	O
costly	O
method	O
of	O
conducting	O
link	B-Application
analysis	I-Application
and	O
has	O
inherent	O
problems	O
of	O
its	O
own	O
.	O
</s>
<s>
Even	O
using	O
domain	O
experts	O
may	O
result	O
in	O
differing	O
conclusions	O
as	O
analysis	B-General_Concept
may	O
be	O
subjective	O
.	O
</s>
<s>
Link	B-Application
analysis	I-Application
techniques	O
have	O
primarily	O
been	O
used	O
for	O
prosecution	O
,	O
as	O
it	O
is	O
far	O
easier	O
to	O
review	O
historical	O
data	O
for	O
patterns	O
than	O
it	O
is	O
to	O
attempt	O
to	O
predict	O
future	O
actions	O
.	O
</s>
<s>
Alternatively	O
,	O
Picarelli	O
argued	O
that	O
use	O
of	O
link	B-Application
analysis	I-Application
techniques	O
could	O
have	O
been	O
used	O
to	O
identify	O
and	O
potentially	O
prevent	O
illicit	O
activities	O
within	O
the	O
Aum	O
Shinrikyo	O
network	O
.	O
</s>
<s>
There	O
are	O
four	O
categories	O
of	O
proposed	O
link	B-Application
analysis	I-Application
solutions	O
:	O
</s>
<s>
Heuristic-based	B-Algorithm
tools	O
utilize	O
decision	O
rules	O
that	O
are	O
distilled	O
from	O
expert	O
knowledge	O
using	O
structured	O
data	O
.	O
</s>
<s>
Template-based	O
tools	O
employ	O
Natural	B-Language
Language	I-Language
Processing	I-Language
(	O
NLP	B-Language
)	O
to	O
extract	O
details	O
from	O
unstructured	B-Application
data	I-Application
that	O
are	O
matched	O
to	O
pre-defined	O
templates	O
.	O
</s>
<s>
J.J.	O
Xu	O
and	O
H	O
.	O
Chen	O
propose	O
a	O
framework	O
for	O
automated	O
network	O
analysis	B-General_Concept
and	O
visualization	B-Application
called	O
CrimeNet	O
Explorer	O
.	O
</s>
<s>
Network	O
Creation	O
through	O
a	O
concept	O
space	O
approach	O
that	O
uses	O
“	O
co-occurrence	B-Application
weight	O
to	O
measure	O
the	O
frequency	O
with	O
which	O
two	O
words	O
or	O
phrases	O
appear	O
in	O
the	O
same	O
document	O
.	O
</s>
<s>
Structural	O
Analysis	B-General_Concept
through	O
“	O
three	O
centrality	O
measures	O
(	O
degree	O
,	O
betweenness	O
,	O
and	O
closeness	O
)	O
to	O
identify	O
central	O
members	O
in	O
a	O
given	O
subgroup	O
.	O
</s>
<s>
CrimeNet	O
Explorer	O
employed	O
Dijkstra	B-Algorithm
’s	I-Algorithm
shortest-path	I-Algorithm
algorithm	I-Algorithm
to	O
calculate	O
the	O
betweenness	O
and	O
closeness	O
from	O
a	O
single	O
node	O
to	O
all	O
other	O
nodes	O
in	O
the	O
subgroup	O
.	O
</s>
<s>
Network	O
Visualization	B-Application
using	O
Torgerson	O
’s	O
metric	O
multidimensional	O
scaling	O
(	O
MDS	O
)	O
algorithm	O
.	O
</s>
