C4.5: programs for machine learning
C4.5: programs for machine learning
Fast discovery of association rules
Advances in knowledge discovery and data mining
Applications of Data Mining to Electronic Commerce
Data Mining and Knowledge Discovery
ICDE '97 Proceedings of the Thirteenth International Conference on Data Engineering
A new approach for measuring rule set consistency
Data & Knowledge Engineering
| Hi-index | 0.01 |
We consider the problem of pruning a given set of if-then rules, such that the support of the pruned rule set is not much less than the support of the given rule set. An empirical measure of similarity between two rules is introduced. This similarity measure is proportional to the degree of overlap between the support sets of the two rules. Using this similarity measure, we cluster the given rule set via the complete linkage algorithm. Rules within a cluster are approximate substitutes for each other and, as such, they can be replaced by a single rule, which is chosen to be the rule whose individual support value is the largest in the cluster. The pruning procedure is demonstrated on a set of rules generated from a marketing data set.