Information Technology Reference
In-Depth Information
9.2 Decision Systems
Action rules and meta-action elementary rules are extracted from decision systems.
A decision system is an information system, where features are partitioned into two
groups: the first group is composed of classification features, and the second of one
particular feature called the decision that models the outcome. An object instance is
represented as a row, called transaction instance, in terms of a set of features.
More formally, by a decision system we mean
S
=
(
X
,
F
∪{
d
}
,
V
)
, where:
1.
X
is a set of objects instances,
F
is a set of classification features,
d
is a decision
feature, and
V
is the domain of these feature values.
2.
f
:
X
ₒ
V
f
is a function for any feature
f
∈
F
, where
V
f
is called the domain
of
f
.
3.
d
:
X
ₒ
V
d
is a function, where
V
d
is called the domain of
d
.
V
d
, where
V
F
=
{
4.
V
=
V
F
∪
V
f
:
f
∈
F
}
.
V
f
is classified
either as positive (normal) or negative (abnormal). To be more precise, we assume
that
F
Also, for each
x
∈
X
and
f
∈
F
, we assume that value
f
(
x
)
∈
(
x
)
denotes the set
{
f
(
x
)
:
f
∈
F
}
which represents the state of the object
instance
x
, and that
F
(
x
)
=
E
n
(
x
)
∪
E
p
(
x
)
, where
E
p
(
x
)
is a set of positive values
and
E
n
(
x
)
is a set of negative values for
x
∈
X
.If
f
(
x
)
∈
E
n
(
x
)
, then the value
f
(
x
)
is interpreted as abnormal (for instance: high temperature, cough, headache).
If
f
(
x
)
∈
E
p
(
x
)
, then value
f
(
x
)
is interpreted as normal.
9.3 Action-Rules
Action rules are rules that provide a set of actionable patterns to follow in order
to transition the objects population from a certain state to a more profitable state
with respect to the decision feature. They allow users to understand the correlations
between transition patterns in the decision system, and construct actionable tasks that
lead to a desirable outcome. Action rules are composed of a decision feature
d
and
classification features that are in turn divided into two sets: stable features
F
st
, and
flexible features
F
fl
such that
F
F
st
. Stable features are object properties that
we do not have control over in the context of our information system. For example,
age and gender are stable features. Flexible features are object properties that can
transition from one value to another value triggering a change in the object state. For
instance, salary and benefits are flexible features since their values can change.
=
F
fl
∪
Definition 1
(
Atomic action term in S
) also called elementary action term in
S
,is
an expression that defines a change of state for a distinct feature in
S
. For example,
(
is an atomic action term which defines a change of value for the
attribute
f
in
S
from
v
1
to
v
2
, where
v
1
,
f
,
v
1
ₒ
v
2
)
V
f
. In the case when there is no
change, we omit the right arrow sign, so for example,
v
2
∈
(
f
,
v
1
)
means that the value of
attribute
f
in
S
remains
v
1
, where
v
1
∈
V
f
.
Search WWH ::
Custom Search