<s>
Association	B-Algorithm
rule	I-Algorithm
learning	I-Algorithm
is	O
a	O
rule-based	B-Algorithm
machine	I-Algorithm
learning	I-Algorithm
method	O
for	O
discovering	O
interesting	O
relations	O
between	O
variables	O
in	O
large	O
databases	O
.	O
</s>
<s>
In	O
any	O
given	O
transaction	O
with	O
a	O
variety	O
of	O
items	O
,	O
association	B-Algorithm
rules	I-Algorithm
are	O
meant	O
to	O
discover	O
the	O
rules	O
that	O
determine	O
how	O
or	O
why	O
certain	O
items	O
are	O
connected	O
.	O
</s>
<s>
Based	O
on	O
the	O
concept	O
of	O
strong	O
rules	O
,	O
Rakesh	O
Agrawal	O
,	O
Tomasz	O
Imieliński	O
and	O
Arun	O
Swami	O
introduced	O
association	B-Algorithm
rules	I-Algorithm
for	O
discovering	O
regularities	O
between	O
products	O
in	O
large-scale	O
transaction	O
data	O
recorded	O
by	O
point-of-sale	O
(	O
POS	O
)	O
systems	O
in	O
supermarkets	O
.	O
</s>
<s>
In	O
addition	O
to	O
the	O
above	O
example	O
from	O
market	B-Algorithm
basket	I-Algorithm
analysis	I-Algorithm
,	O
association	B-Algorithm
rules	I-Algorithm
are	O
employed	O
today	O
in	O
many	O
application	O
areas	O
including	O
Web	B-Application
usage	I-Application
mining	I-Application
,	O
intrusion	O
detection	O
,	O
continuous	O
production	O
,	O
and	O
bioinformatics	O
.	O
</s>
<s>
In	O
contrast	O
with	O
sequence	B-Algorithm
mining	I-Algorithm
,	O
association	B-Algorithm
rule	I-Algorithm
learning	I-Algorithm
typically	O
does	O
not	O
consider	O
the	O
order	O
of	O
items	O
either	O
within	O
a	O
transaction	O
or	O
across	O
transactions	O
.	O
</s>
<s>
The	O
association	B-Algorithm
rule	I-Algorithm
algorithm	O
itself	O
consists	O
of	O
various	O
parameters	O
that	O
can	O
make	O
it	O
difficult	O
for	O
those	O
without	O
some	O
expertise	O
in	O
data	B-Application
mining	I-Application
to	O
execute	O
,	O
with	O
many	O
rules	O
that	O
are	O
arduous	O
to	O
understand	O
.	O
</s>
<s>
Following	O
the	O
original	O
definition	O
by	O
Agrawal	O
,	O
Imieliński	O
,	O
Swami	O
the	O
problem	O
of	O
association	B-Algorithm
rule	I-Algorithm
mining	I-Algorithm
is	O
defined	O
as	O
:	O
</s>
<s>
Association	B-Algorithm
rules	I-Algorithm
are	O
made	O
by	O
searching	O
data	O
for	O
frequent	O
if-then	O
patterns	O
and	O
by	O
using	O
a	O
certain	O
criterion	O
under	O
Support	O
and	O
Confidence	O
to	O
define	O
what	O
the	O
most	O
important	O
relationships	O
are	O
.	O
</s>
<s>
However	O
,	O
there	O
is	O
a	O
third	O
criteria	O
that	O
can	O
be	O
used	O
,	O
it	O
is	O
called	O
Lift	B-Algorithm
and	O
it	O
can	O
be	O
used	O
to	O
compare	O
the	O
expected	O
Confidence	O
and	O
the	O
actual	O
Confidence	O
.	O
</s>
<s>
Lift	B-Algorithm
will	O
show	O
how	O
many	O
times	O
the	O
if-then	O
statement	O
is	O
expected	O
to	O
be	O
found	O
to	O
be	O
true	O
.	O
</s>
<s>
Association	B-Algorithm
rules	I-Algorithm
are	O
made	O
to	O
calculate	O
from	O
itemsets	O
,	O
which	O
are	O
created	O
by	O
two	O
or	O
more	O
items	O
.	O
</s>
<s>
That	O
is	O
why	O
Association	B-Algorithm
rules	I-Algorithm
are	O
typically	O
made	O
from	O
rules	O
that	O
are	O
well	O
represented	O
by	O
the	O
data	O
.	O
</s>
<s>
There	O
are	O
many	O
different	O
data	B-Application
mining	I-Application
techniques	O
you	O
could	O
use	O
to	O
find	O
certain	O
analytics	O
and	O
results	O
,	O
for	O
example	O
,	O
there	O
is	O
Classification	O
analysis	O
,	O
Clustering	O
analysis	O
,	O
and	O
Regression	O
analysis	O
.	O
</s>
<s>
Association	B-Algorithm
rules	I-Algorithm
are	O
primarily	O
used	O
to	O
find	O
analytics	O
and	O
a	O
prediction	O
of	O
customer	O
behavior	O
.	O
</s>
<s>
There	O
are	O
many	O
benefits	O
of	O
using	O
Association	B-Algorithm
rules	I-Algorithm
like	O
finding	O
the	O
pattern	O
that	O
helps	O
understand	O
the	O
correlations	O
and	O
co-occurrences	O
between	O
data	O
sets	O
.	O
</s>
<s>
A	O
very	O
good	O
real-world	O
example	O
that	O
uses	O
Association	B-Algorithm
rules	I-Algorithm
would	O
be	O
medicine	O
.	O
</s>
<s>
Medicine	O
uses	O
Association	B-Algorithm
rules	I-Algorithm
to	O
help	O
diagnose	O
patients	O
.	O
</s>
<s>
With	O
the	O
use	O
of	O
the	O
Association	B-Algorithm
rules	I-Algorithm
,	O
doctors	O
can	O
determine	O
the	O
conditional	O
probability	O
of	O
an	O
illness	O
by	O
comparing	O
symptom	O
relationships	O
from	O
past	O
cases	O
.	O
</s>
<s>
However	O
,	O
Association	B-Algorithm
rules	I-Algorithm
also	O
lead	O
to	O
many	O
different	O
downsides	O
such	O
as	O
finding	O
the	O
appropriate	O
parameter	O
and	O
threshold	O
settings	O
for	O
the	O
mining	O
algorithm	O
.	O
</s>
<s>
For	O
someone	O
that	O
does	O
n’t	O
have	O
a	O
good	O
concept	O
of	O
data	B-Application
mining	I-Application
,	O
this	O
might	O
cause	O
them	O
to	O
have	O
trouble	O
understanding	O
it	O
.	O
</s>
<s>
ThresholdsWhen	O
using	O
Association	B-Algorithm
rules	I-Algorithm
,	O
you	O
are	O
most	O
likely	O
to	O
only	O
use	O
Support	O
and	O
Confidence	O
.	O
</s>
<s>
Usually	O
,	O
the	O
Association	B-Algorithm
rule	I-Algorithm
generation	O
is	O
split	O
into	O
two	O
different	O
steps	O
that	O
needs	O
to	O
be	O
applied	O
:	O
</s>
<s>
Exploiting	O
this	O
property	O
,	O
efficient	O
algorithms	O
(	O
e.g.	O
,	O
Apriori	B-Algorithm
and	O
Eclat	O
)	O
can	O
find	O
all	O
frequent	O
itemsets	O
.	O
</s>
<s>
Let	O
be	O
itemsets	O
,	O
an	O
association	B-Algorithm
rule	I-Algorithm
and	O
a	O
set	O
of	O
transactions	O
of	O
a	O
given	O
database	O
.	O
</s>
<s>
When	O
using	O
antecedents	O
and	O
consequents	O
,	O
it	O
allows	O
a	O
data	B-Application
miner	I-Application
to	O
determine	O
the	O
support	O
of	O
multiple	O
items	O
being	O
bought	O
together	O
in	O
comparison	O
to	O
the	O
whole	O
data	O
set	O
.	O
</s>
<s>
Another	O
way	O
of	O
finding	O
interesting	O
samples	O
is	O
to	O
find	O
the	O
value	O
of	O
(	O
support	O
)	O
X(confidence )	O
;	O
this	O
allows	O
a	O
data	B-Application
miner	I-Application
to	O
see	O
the	O
samples	O
where	O
support	O
and	O
confidence	O
are	O
high	O
enough	O
to	O
be	O
highlighted	O
in	O
the	O
dataset	O
and	O
prompt	O
a	O
closer	O
look	O
at	O
the	O
sample	O
to	O
find	O
more	O
information	O
on	O
the	O
connection	O
between	O
the	O
items	O
.	O
</s>
<s>
With	O
respect	O
to	O
,	O
the	O
confidence	O
value	O
of	O
an	O
association	B-Algorithm
rule	I-Algorithm
,	O
often	O
denoted	O
as	O
,	O
is	O
the	O
ratio	O
of	O
transactions	O
containing	O
both	O
and	O
to	O
the	O
total	O
amount	O
of	O
values	O
present	O
,	O
where	O
is	O
the	O
antecedent	O
and	O
is	O
the	O
consequent	O
.	O
</s>
<s>
Table	O
4	O
shows	O
association	B-Algorithm
rule	I-Algorithm
examples	O
where	O
the	O
minimum	O
threshold	O
for	O
confidence	O
is	O
0.5	O
(	O
50%	O
)	O
.	O
</s>
<s>
Overall	O
,	O
using	O
confidence	O
in	O
association	B-Algorithm
rule	I-Algorithm
mining	I-Algorithm
is	O
great	O
way	O
to	O
bring	O
awareness	O
to	O
data	O
relations	O
.	O
</s>
<s>
However	O
,	O
confidence	O
is	O
not	O
the	O
optimal	O
method	O
for	O
every	O
concept	O
in	O
association	B-Algorithm
rule	I-Algorithm
mining	I-Algorithm
.	O
</s>
<s>
The	O
lift	B-Algorithm
of	O
a	O
rule	O
is	O
defined	O
as	O
:	O
</s>
<s>
For	O
example	O
,	O
the	O
rule	O
has	O
a	O
lift	B-Algorithm
of	O
.	O
</s>
<s>
If	O
the	O
rule	O
had	O
a	O
lift	B-Algorithm
of	O
1	O
,	O
it	O
would	O
imply	O
that	O
the	O
probability	O
of	O
occurrence	O
of	O
the	O
antecedent	O
and	O
that	O
of	O
the	O
consequent	O
are	O
independent	O
of	O
each	O
other	O
.	O
</s>
<s>
If	O
the	O
lift	B-Algorithm
is	O
>	O
1	O
,	O
that	O
lets	O
us	O
know	O
the	O
degree	O
to	O
which	O
those	O
two	O
occurrences	O
are	O
dependent	O
on	O
one	O
another	O
,	O
and	O
makes	O
those	O
rules	O
potentially	O
useful	O
for	O
predicting	O
the	O
consequent	O
in	O
future	O
data	O
sets	O
.	O
</s>
<s>
If	O
the	O
lift	B-Algorithm
is	O
<	O
1	O
,	O
that	O
lets	O
us	O
know	O
the	O
items	O
are	O
substitute	O
to	O
each	O
other	O
.	O
</s>
<s>
The	O
value	O
of	O
lift	B-Algorithm
is	O
that	O
it	O
considers	O
both	O
the	O
support	O
of	O
the	O
rule	O
and	O
the	O
overall	O
data	O
set	O
.	O
</s>
<s>
The	O
concept	O
of	O
association	B-Algorithm
rules	I-Algorithm
was	O
popularized	O
particularly	O
due	O
to	O
the	O
1993	O
article	O
of	O
Agrawal	O
et	O
al.	O
,	O
which	O
has	O
acquired	O
more	O
than	O
23,790	O
citations	O
according	O
to	O
Google	O
Scholar	O
,	O
as	O
of	O
April	O
2021	O
,	O
and	O
is	O
thus	O
one	O
of	O
the	O
most	O
cited	O
papers	O
in	O
the	O
Data	B-Application
Mining	I-Application
field	O
.	O
</s>
<s>
However	O
,	O
what	O
is	O
now	O
called	O
"	O
association	B-Algorithm
rules	I-Algorithm
"	O
is	O
introduced	O
already	O
in	O
the	O
1966	O
paper	O
on	O
GUHA	O
,	O
a	O
general	O
data	B-Application
mining	I-Application
method	O
developed	O
by	O
Petr	O
Hájek	O
et	O
al	O
.	O
</s>
<s>
An	O
early	O
(	O
circa	O
1989	O
)	O
use	O
of	O
minimum	O
support	O
and	O
confidence	O
to	O
find	O
all	O
association	B-Algorithm
rules	I-Algorithm
is	O
the	O
Feature	O
Based	O
Modeling	O
framework	O
,	O
which	O
found	O
all	O
rules	O
with	O
and	O
greater	O
than	O
user	O
defined	O
constraints	O
.	O
</s>
<s>
Statistically	O
sound	O
association	B-Algorithm
discovery	I-Algorithm
controls	O
this	O
risk	O
,	O
in	O
most	O
cases	O
reducing	O
the	O
risk	O
of	O
finding	O
any	O
spurious	O
associations	O
to	O
a	O
user-specified	O
significance	O
level	O
.	O
</s>
<s>
Many	O
algorithms	O
for	O
generating	O
association	B-Algorithm
rules	I-Algorithm
have	O
been	O
proposed	O
.	O
</s>
<s>
Some	O
well-known	O
algorithms	O
are	O
Apriori	B-Algorithm
,	O
Eclat	O
and	O
FP-Growth	O
,	O
but	O
they	O
only	O
do	O
half	O
the	O
job	O
,	O
since	O
they	O
are	O
algorithms	O
for	O
mining	O
frequent	O
itemsets	O
.	O
</s>
<s>
Apriori	B-Algorithm
is	O
given	O
by	O
R	O
.	O
Agrawal	O
and	O
R	O
.	O
Srikant	O
in	O
1994	O
for	O
frequent	O
item	O
set	O
mining	O
and	O
association	B-Algorithm
rule	I-Algorithm
learning	I-Algorithm
.	O
</s>
<s>
The	O
name	O
of	O
the	O
algorithm	O
is	O
Apriori	B-Algorithm
because	O
it	O
uses	O
prior	O
knowledge	O
of	O
frequent	O
itemset	O
properties	O
.	O
</s>
<s>
Overview	O
:	O
Apriori	B-Algorithm
uses	O
a	O
"	O
bottom	O
up	O
"	O
approach	O
,	O
where	O
frequent	O
subsets	O
are	O
extended	O
one	O
item	O
at	O
a	O
time	O
(	O
a	O
step	O
known	O
as	O
candidate	O
generation	O
)	O
,	O
and	O
groups	O
of	O
candidates	O
are	O
tested	O
against	O
the	O
data	O
.	O
</s>
<s>
Apriori	B-Algorithm
uses	O
breadth-first	B-Algorithm
search	I-Algorithm
and	O
a	O
Hash	B-Application
tree	I-Application
structure	O
to	O
count	O
candidate	O
item	O
sets	O
efficiently	O
.	O
</s>
<s>
Apriori	B-Algorithm
has	O
some	O
limitations	O
.	O
</s>
<s>
Apriori	B-Algorithm
is	O
slower	O
than	O
the	O
Eclat	O
algorithm	O
.	O
</s>
<s>
However	O
,	O
Apriori	B-Algorithm
performs	O
well	O
compared	O
to	O
Eclat	O
when	O
the	O
dataset	O
is	O
large	O
.	O
</s>
<s>
FP-growth	O
outperforms	O
the	O
Apriori	B-Algorithm
and	O
Eclat	O
.	O
</s>
<s>
ECLAT	O
,	O
stands	O
for	O
Equivalence	O
Class	O
Transformation	O
)	O
is	O
a	O
backtracking	B-Algorithm
algorithm	I-Algorithm
,	O
which	O
traverses	O
the	O
frequent	O
itemset	O
lattice	O
graph	O
in	O
a	O
depth-first	B-Algorithm
search	I-Algorithm
(	O
DFS	O
)	O
fashion	O
.	O
</s>
<s>
Whereas	O
the	O
breadth-first	B-Algorithm
search	I-Algorithm
(	O
BFS	O
)	O
traversal	O
used	O
in	O
the	O
Apriori	B-Algorithm
algorithm	I-Algorithm
will	O
end	O
up	O
checking	O
every	O
subset	O
of	O
an	O
itemset	O
before	O
checking	O
it	O
,	O
DFS	O
traversal	O
checks	O
larger	O
itemsets	O
and	O
can	O
save	O
on	O
checking	O
the	O
support	O
of	O
some	O
of	O
its	O
subsets	O
by	O
virtue	O
of	O
the	O
downward-closer	O
property	O
.	O
</s>
<s>
In	O
the	O
second	O
pass	O
,	O
it	O
builds	O
the	O
FP-tree	O
structure	O
by	O
inserting	O
transactions	O
into	O
a	O
trie	B-General_Concept
.	O
</s>
<s>
Recursive	O
processing	O
of	O
this	O
compressed	O
version	O
of	O
the	O
main	O
dataset	O
grows	O
frequent	O
item	O
sets	O
directly	O
,	O
instead	O
of	O
generating	O
candidate	O
items	O
and	O
testing	O
them	O
against	O
the	O
entire	O
database	O
(	O
as	O
in	O
the	O
apriori	B-Algorithm
algorithm	I-Algorithm
)	O
.	O
</s>
<s>
Once	O
the	O
recursive	O
process	O
has	O
completed	O
,	O
all	O
frequent	O
item	O
sets	O
will	O
have	O
been	O
found	O
,	O
and	O
association	B-Algorithm
rule	I-Algorithm
creation	O
begins	O
.	O
</s>
<s>
The	O
ASSOC	O
procedure	O
is	O
a	O
GUHA	O
method	O
which	O
mines	O
for	O
generalized	O
association	B-Algorithm
rules	I-Algorithm
using	O
fast	O
bitstrings	B-Data_Structure
operations	O
.	O
</s>
<s>
The	O
association	B-Algorithm
rules	I-Algorithm
mined	O
by	O
this	O
method	O
are	O
more	O
general	O
than	O
those	O
output	O
by	O
apriori	B-Algorithm
,	O
for	O
example	O
"	O
items	O
"	O
can	O
be	O
connected	O
both	O
with	O
conjunction	O
and	O
disjunctions	O
and	O
the	O
relation	O
between	O
antecedent	O
and	O
consequent	O
of	O
the	O
rule	O
is	O
not	O
restricted	O
to	O
setting	O
minimum	O
support	O
and	O
confidence	O
as	O
in	O
apriori	B-Algorithm
:	O
an	O
arbitrary	O
combination	O
of	O
supported	O
interest	O
measures	O
can	O
be	O
used	O
.	O
</s>
<s>
OPUS	O
is	O
an	O
efficient	O
algorithm	O
for	O
rule	B-Algorithm
discovery	I-Algorithm
that	O
,	O
in	O
contrast	O
to	O
most	O
alternatives	O
,	O
does	O
not	O
require	O
either	O
monotone	O
or	O
anti-monotone	O
constraints	O
such	O
as	O
minimum	O
support	O
.	O
</s>
<s>
OPUS	O
search	O
is	O
the	O
core	O
technology	O
in	O
the	O
popular	O
Magnum	O
Opus	O
association	B-Algorithm
discovery	I-Algorithm
system	O
.	O
</s>
<s>
A	O
famous	O
story	O
about	O
association	B-Algorithm
rule	I-Algorithm
mining	I-Algorithm
is	O
the	O
"	O
beer	O
and	O
diaper	O
"	O
story	O
.	O
</s>
<s>
This	O
anecdote	O
became	O
popular	O
as	O
an	O
example	O
of	O
how	O
unexpected	O
association	B-Algorithm
rules	I-Algorithm
might	O
be	O
found	O
from	O
everyday	O
data	O
.	O
</s>
<s>
Multi-Relation	O
Association	B-Algorithm
Rules	I-Algorithm
(	O
MRAR	O
)	O
:	O
These	O
are	O
association	B-Algorithm
rules	I-Algorithm
where	O
each	O
item	O
may	O
have	O
several	O
relations	O
.	O
</s>
<s>
Such	O
association	B-Algorithm
rules	I-Algorithm
can	O
be	O
extracted	O
from	O
RDBMS	O
data	O
or	O
semantic	O
web	O
data	O
.	O
</s>
<s>
Contrast	B-Algorithm
set	I-Algorithm
learning	I-Algorithm
is	O
a	O
form	O
of	O
associative	O
learning	O
.	O
</s>
<s>
Weighted	O
class	O
learning	O
is	O
another	O
form	O
of	O
associative	O
learning	O
where	O
weights	O
may	O
be	O
assigned	O
to	O
classes	O
to	O
give	O
focus	O
to	O
a	O
particular	O
issue	O
of	O
concern	O
for	O
the	O
consumer	O
of	O
the	O
data	B-Application
mining	I-Application
results	O
.	O
</s>
<s>
K-optimal	B-Algorithm
pattern	I-Algorithm
discovery	I-Algorithm
provides	O
an	O
alternative	O
to	O
the	O
standard	O
approach	O
to	O
association	B-Algorithm
rule	I-Algorithm
learning	I-Algorithm
which	O
requires	O
that	O
each	O
pattern	O
appear	O
frequently	O
in	O
the	O
data	O
.	O
</s>
<s>
Approximate	O
Frequent	O
Itemset	B-Algorithm
mining	I-Algorithm
is	O
a	O
relaxed	O
version	O
of	O
Frequent	O
Itemset	B-Algorithm
mining	I-Algorithm
that	O
allows	O
some	O
of	O
the	O
items	O
in	O
some	O
of	O
the	O
rows	O
to	O
be	O
0	O
.	O
</s>
<s>
Interval	O
Data	O
Association	B-Algorithm
Rules	I-Algorithm
e.g.	O
</s>
<s>
Sequential	B-Algorithm
pattern	I-Algorithm
mining	I-Algorithm
discovers	O
subsequences	O
that	O
are	O
common	O
to	O
more	O
than	O
minsup	O
(	O
minimum	O
support	O
threshold	O
)	O
sequences	O
in	O
a	O
sequence	O
database	O
,	O
where	O
minsup	O
is	O
set	O
by	O
the	O
user	O
.	O
</s>
<s>
Subspace	O
Clustering	O
,	O
a	O
specific	O
type	O
of	O
clustering	B-Algorithm
high-dimensional	I-Algorithm
data	I-Algorithm
,	O
is	O
in	O
many	O
variants	O
also	O
based	O
on	O
the	O
downward-closure	O
property	O
for	O
specific	O
clustering	O
models	O
.	O
</s>
<s>
Warmr	O
,	O
shipped	O
as	O
part	O
of	O
the	O
ACE	O
data	B-Application
mining	I-Application
suite	O
,	O
allows	O
association	B-Algorithm
rule	I-Algorithm
learning	I-Algorithm
for	O
first	O
order	O
relational	O
rules	O
.	O
</s>
