Databases Reference
In-Depth Information
propose E-action rules that enhance the extended action rules by adding to
its descriptions the values of common stable attributes listed in both clas-
sification rules, used to construct an action rule, in order to provide a more
sound and well-defined strategy. Any E-action rule provides a well defined hint
to a user of what changes within flexible attributes are needed to re-classify
some objects from a lower preferable class to a higher one. Hence, E-action
rules mining is a technique that intelligently and automatically forms precise
actions that can be adopted by decision-making users in achieving their goals.
System
DEAR-2
initially generates a set of classification rules from
S
(sat-
isfying two thresholds, the first one for a minimum support and second for
a minimum confidence) defining values of a chosen attribute, called decision
attribute in
S
, in terms of the remaining attributes.
DEAR-2
is giving pref-
erence to rules which classification part contains maximally small number of
stable attributes in
S
. These rules are partitioned by
DEAR-2
into a number
of equivalence classes where each equivalence class contains only rules which
classification part has the same values of stable attributes. Each equivalence
class is used independently by
DEAR-2
as a base for constructing action rules.
The current strategy requires the generation of classification rules from
S
to
form a base, before the process of action rules construction starts. We believe
that by following the process similar to
LERS
(see [2, 5]) or
ERID
(see [3])
which is initially centered on all stable attributes in
S
, we should be able to
construct action rules directly from
S
and without the necessity to generate
the base of classification rules.
References
1. Adomavicius G, Tuzhilin A (1997) Discovery of actionable patterns in data-
bases: the action hierarchy approach. In: Proceedings of KDD97 Conference.
AAAI, Newport Beach, CA
2. Chmielewski M R, Grzymala-Busse J W, Peterson N W, Than S (1993) The
rule induction system LERS - a version for personal computers. In: Foundations
of Computing and Decision Sciences. Vol. 18, No. 3-4, Institute of Computing
Science, Technical University of Poznan, Poland, 181-212
3. Dardzinska A, Ras Z W (2003) On rule discovery from incomplete informa-
tion systems. In: Proceedings of ICDM'03 Workshop on Foundations and New
Directions of Data Mining, (Eds: Lin T Y, Hu X, Ohsuga S, Liau C). IEEE
Computer Society, Melbourne, Florida, 31-35
4. Geffner H, Wainer J (1998) Modeling action, knowledge and control. In: ECAI
98, Proceedings of the 13th European Conference on AI, (Ed: Prade H). Wiley,
New York, 532-536
5. Grzymala-Busse J (1997) A new version of the rule induction system LERS.
In: Fundamenta Informaticae, Vol. 31, No. 1, 27-39
6. Liu B, Hsu W, Chen S (1997) Using general impressions to analyze discov-
ered classification rules. In: Proceedings of KDD97 Conference. AAAI, Newport
Beach, CA
