Information Technology Reference
In-Depth Information
In this context, the support of a classification rule
r
=
[
[
(
f
1
,
f
11
)
∧
(
f
2
,
f
21
)
∧
··· ∧
(
f
k
,
f
k
1
)
]ₒ
(
,
d
1
)
=
(
,
∪{
}
,
)
(
∀
≤
d
] in a decision system
S
X
F
d
V
, where
i
k
)(
f
i
∈
F
and
f
i
1
∈
V
f
i
)
,
d
1
∈
V
d
is defined as
sup
(
r
)
={
x
∈
X
:
(
∀
i
≤
k
)
[
f
i
(
x
)
=
f
i
1
]
and
d
(
x
)
=
d
1
}
.
9.4 Meta-actions
Action rules are the perfect analysis of transition patterns that inform deciders about
the possible changes to perform to reach a desired outcome. However, deciders still
need to acquire additional knowledge on how to perform the necessary changes, and
what are the object's states changes that actually occurred in the system. To build
the strategies on the top of the actionable tasks that action rules provide, we use
meta-actions that are defined as follows.
Definition 4
(
Meta-actions
) associated with a decision system
S
are defined as
higher level concepts used to model certain generalizations of actions rules [
14
].
Meta-actions, when executed, trigger changes in values of some flexible features in
S
.
Meta-actions are actions, outside of the features
F
, taken by deciders to transition
objects from an initial known state with specific preconditions to different state with
known postconditions. The changes in flexible features, triggered by meta-actions,
are represented by atomic action terms for the respective features, and reported by
the influence matrix presented in [
14
].
Example 1
Let us take market segmentation for automobiles as an example of
domain. Then, an automotive company, say company
X
, would divide its customers
into: sedan cars seekers, sport cars seekers, wagon car seekers, all roads car seekers,
and hatchback cars seekers. An extensive list of classification features can be used
to classify them. For instance, age would be a good feature that would inform us that
a young person would prefer a hatchback, and an older person would rather have a
sedan. Another feature such as number of kids would inform us that bigger families
would prefer wagons or all roads rather than smaller cars, marital status could be a
good indication of sport cars preference, and so on. Other feature can be analyzed
for the purpose of customer satisfaction and segmentation. However, all features
cited earlier are stable features and would not allow values transitions regardless of
meta-actions applied.
Let us assume another company
Y
, a luxurious car company, would like to acquire
new customers, then their market segmentation would differ from
X
's market seg-
mentation. In fact, the new segmentation would be based on new classification fea-
tures, some of them are: customers income, car price range, customer functional
needs, car comfort, car quality, and customers favorite brand. Given those features,
the company can classify the customers based on their favorite brand, and would like
to attract customers from other less luxurious brands. Based on those features,
Y
can
apply a meta-action
M
that transitions car price range to match what customers can
Search WWH ::
Custom Search