A new technique for fast frequent closed itemsets mining

This paper proposes a new technique for fast frequent closed itemsets mining. The method employs the concept of prefix-based equivalence class to partition the database in such a way that the possible frequent itemsets will be uniformly distributed in the partitions. One unique property of this technique is that it guarantees only frequent closed itemsets are generated and they are generated only once. Thus it avoids the closed-set examination in the entire mining processing. Our technique can be used in the existing frequent closed itemsets mining algorithms, including both Apriori-based and frequent pattern growth based algorithms, to improve their performance. Moreover, the technique is well suited for parallel mining of frequent closed itemsets.