Information Technology Reference
In-Depth Information
Usually the following four parameters are applied to describe the attributes of
rules
(1) Confidence
Suppose c% transactions which support itemset
A
in
W
also support itemset
B
at the same time, then
c
% is the confidence of association rule
A
¼
B
. That is,
confidence stands for the probability which a itemse
A
occurs in the transaction
T
and a itemset
B
also occurs in transaction
T
. The rule
A
¼
B
has confidence
c
in
the transaction set
W
, where
c
is the percentage of transactions in
W
containing
A
that also contain
B
. This is taken to be the conditional probability,
P
(
B|A
). That
is,
confidence(
). (12.1)
For example, suppose 70% consummer who bought bread also bought butter,
then the confidence is 70%.
(2) Support
Suppose
A
¼
B
) =
P
(
B|A
s
% transactions in
W
supporting both the itemset
A
and
B
,
s
% is
called the support of rule
A
¼
B
. The rule
A
¼
B
holds in the transaction set
W
with support
s
, where
s
is the percentage of transactions in
W
that contain
A B
,
that is, the union of set
A
and
B
. This is taken to be the probability,
P
(
A B
)
which indicates a transaction contains the union of set
.
For example, suppose there are 1000 customers in a supermarket someday
and among them 100 customers purchased both bread and butter, then the
support of
A
and set
B
is 10%(100/1000).
(3) Expected Confidence
Suppose
A
¼
B
e
% transactions in
W
support the itemset
B
, then e% is called the
expected confidence of rule
A
¼
B
. Expected confidence reflects the probability
of the itemset
occurs in all transactions when there are no other restrictions.
For example, suppose there are 1000 customersr in a supermarket and 200
customers purchased butter, then the expected confidence of
B
A
¼
B
is 20%.
(4) Lift
Lift is the ratio of confidence to the expected confidence. Lift reflects the
influence which the occurrence of itemsets
A
to the occurrence of itemset
B
.
Search WWH ::




Custom Search