Databases Reference
In-Depth Information
and subjective (see [1, 6, 13, 14]). Subjective measure is when the judgment is
made by people and objective measure is basically when the judgment is made
by computer derived methods or strategies. Generally, they evaluate the rules
based on their quality and similarity between them.
A rule is deemed actionable, if users can take action to gain an advantage
based on that rule [6]. This definition, in spite of its importance, is too vague
and it leaves open door to a number of different interpretations of action-
ability. In order to narrow it down, a new class of rules (called action rules)
constructed from certain pairs of classification rules, has been proposed in [10].
A formal definition of an action rule was independently proposed in [4]. These
rules have been investigated further in [11].
Classical action rules involve only flexible attributes listed in both clas-
sification rules from which the action rule is constructed. Extended action
rule is a significant improvement of a classical action rule because of the con-
straints placed on values of attributes listed only in one of these rules. But,
still extended action rules do not include the values of common stable at-
tributes listed in rules from which they are constructed. This implies that the
domain of an action rule does not reflect correctly to what class of objects it
can be successfully applied. To solve this problem, a new class of rules, called
E-Action rules, is proposed. E-action rules extract actionability knowledge
among pairs of classification rules in a more accurate way.
E-action rules are useful is many fields, including medical diagnosis and
business. In medical diagnosis, classification rules can explain the relationships
between symptoms and type of sickness as well as predict the diagnosis of a
new patient. Extended action rules are useful in providing suggestions for
modifying some symptoms in order to recover from an illness. In business,
classification rules can distinguish the good customers from the bad ones.
E-action rules can provide specific actions that can be taken by decision-
makers to re-classify customers.
The strategy for generating action rules proposed in [11] is significantly
improved in the system
DEAR-2
presented in this chapter. It consists of three
steps. The first step is to partition the rules into equivalence classes with
respect to the values of the decision attribute. In the second step, we use a
recursive algorithm to dynamically build a tree structure partitioning each
class of rules based on values of their stable attributes. When any two rules
have the same stable attribute values, they are placed in the same class. In the
final step, instead of comparing all pairs of rules, only pairs of rules belonging
to some of these equivalent classes have to be compared in order to construct
extended action rules. This strategy significantly reduces the number of steps
needed to generate action rules in comparison to the strategy (called
DEAR
)
proposed in [11].
In this chapter, we present a new definition of E-Action rules to enhance
the action rule and the extended action rules.
