Discovering Frequent Pattern Pairs
Cubes and association rules discover frequent patterns in a data set, most of which are not significant. Thus previous research has introduced search constraints and statistical metrics to discover significant patterns and reduce processing time. The authors introduce cube pairs (comparing cube groups based on a parametric statistical test) and rule pairs (based on two similar association rules), which are pattern pair generalizations of cubes and association rules, respectively. They introduce algorithmic optimizations to discover comparable pattern sets. They carefully study why both techniques agree or disagree on the validity of specific pairs, considering p-value for statistical tests, as well as confidence for association rules.