Post-mining: Maintenance of association rules by weighting

This paper proposes a new strategy for maintaining association rules in dynamic databases. This method uses weighting technique to highlight new data. Our approach is novel in that recently added transactions are given higher weights. In particular, we look at how frequent itemsets can be maintained incrementally. We propose a competitive model to 'promote' infrequent itemsets to frequent itemsets, and to 'degrade' frequent itemsets to infrequent itemsets incrementally. This competitive strategy can avoid retracing the whole data set. We have evaluated the proposed method. The experiments have shown that our approach is efficient and promising.