An Algorithm for Mining Maximum Frequent Itemsets Using Data-sets Condensing and Intersection Pruning

Discovering maximal frequent itemset is a key issue in data mining; the Apriori-like algorithms use candidate itemsets generating/testing method, but this approach is highly time-consuming. To look for an algorithm that can avoid the generating of vast volume of candidate itemsets, nor the generating of frequent pattern tree, DCIP algorithm uses data-set condensing and intersection pruning to find the maximal frequent itemset. The condensing process is performed by deleting items in infrequent 1-itemset and merging duplicate transactions repeatedly; the pruning process is performed by generating intersections of transactions and deleting unneeded subsets recursively. This algorithm differs from all classical maximal frequent itemset discovering algorithms; experiments show that this algorithm is valid with moderate efficiency; it is also easy to code for use in KDD applications.