Authors:
Xuan-Hiep Huynh
;
Fabrice Guillet
and
Henri Briand
Affiliation:
LINA FRE CNRS 2729 - Polytechnic school of Nantes university, France
Keyword(s):
interestingness measure, stable cluster, post-processing, association rules, knowledge quality.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Biomedical Engineering
;
Business Analytics
;
Data Engineering
;
Data Mining
;
Databases and Information Systems Integration
;
Datamining
;
Enterprise Information Systems
;
Health Information Systems
;
Sensor Networks
;
Signal Processing
;
Soft Computing
Abstract:
In this paper, dealing with association rules post-processing, we propose to study the correlations between 36 interestingness measures (IM), in order to better understand their behavior on data and finally to help the data miner chooses the best IMs. We used two datasets with opposite characteristics in which we extract two rulesets about 100000 rules, and the two subsets of the 1000 best rules according to IMs. The study of the correlation between IMs with PAM and AHC shows unexpected stabilities between the four ruleset, and more precisely eight stable clusters of IMs are found and described.