Meta-actions as a Tool for Action Rules Evaluation - Feature Selection for Data and Pattern Recognition - page 190

replacement; removal of cardiac pacemaker or cardioverter/defibrillator, CCS code
of 48. However, depending of the cost of the meta-action and the satisfaction rate, a
practitioner may make a different decision.
Similarly, we can apply meta-actions M

} ={

44 } , Coronary artery bypass graft

{ 44

(CABG), to trigger r 2 and resolve the diagnoses Coagulation and

hemorrhagic disorders, Fluid and electrolyte disorders, Cardiac dysrhythmias} and
we will get: GlobConf

{ 62 , 55 , 106 }={

32%.
We lack the cost of meta-actions in our dataset; hence, we cannot compute the
SatRate . Nonetheless, this information can be obtained via consultation with a prac-
titioner. If we assume that the cost of any given meta-action is the same and that

( r 1 ) = 94% and ExConf ( M

} (

r 1 )) = 73 .

{ 48

λ
is selected as a constant for each meta-action, then the SatRate will essentially be
equal to the ExConf and practitioners can base their decision on the best execution
confidence.

9.8 Conclusion

Nowadays, action rules are used in several industries and the healthcare industry
among others is a very sensitive area. Results from action rules extraction process
have to be thoroughly evaluated and analyzed to be used in such industries. Action
rules are commonly constructed from feature values patterns and not from transition
patterns. In this chapter, we used meta-actions to evaluate action rules and we intro-
duced new evaluation metrics. We used the 2010 Florida State Inpatient Databases
(SID), and extracted meta-actions and action rules from this dataset. We evaluated
meta-actions applied to action rules with the different metrics and compared the
results to traditional metrics. Our results show the effectiveness of meta-actions in
evaluating action rules and the rigorousness of our evaluation metrics. In future work,
we will explore the meta-actions' extra action terms, considered as side effects, that
covered patients' preconditions and did not cover action rules when applying meta-
actions.

References

1. Agrawal, R., Srikant, R.: Fast algorithms for mining association rules in large databases.
In: Proceedings of the 20th International Conference on Very Large Data Bases. VLDB'94,
pp. 487-499. Morgan Kaufmann Publishers Inc., San Francisco (1994)
2. Cost, H., (HCUP), U.P.: HCUP state inpatient databases (SID), agency for healthcare research
and quality, rockville, md. www.hcup-us.ahrq.gov/sidoverview.jsp (2005-2009)
3. Cost, H., (HCUP), U.P., for Healthcare Research, A., Quality: Clinical classifications software
(CCS) for ICD-9-CM. Website. http://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp
4. Im, S., Ras, Z.: Action rule extraction from a decision table: ARED. In: Foundations of Intel-
ligent Systems. Proceedings of ISMIS'08, pp. 160-168. Springer, Toronto (2008)