A Galois Lattice framework to handle updates in the mining of closed itemsets in dynamic databases
Compute '08: Proceedings of the 1st Bangalore annual Compute conference, page 1--6. New York, NY, USA, ACM, (2008)DOI: http://doi.acm.org/10.1145/1341771.1341788
Abstract
Incrementality is a major challenge in the mining of dynamic databases. In such databases, the maintenance of association rules can be directly mapped into the problem of maintaining closed frequent itemsets. A number of incremental strategies have been proposed earlier with several limitations. A serious limitation is the need to examine the entire family of closed itemsets, whenever there are insertions or deletions in the database. The proposed strategy relies on an efficient and selective update of the closed itemsets using an indexed trie structure. The framework emphasizes on certain fundamental and structural properties of Galois Lattice theory to overcome the limitations of the earlier approaches. Apart from facilitating a selective update, the indexed structure removes the necessity of working with a wholly memory resident trie.
Description
A Galois Lattice framework to handle updates in the mining of closed itemsets in dynamic databases
