Domain and data partitioning for parallel mining of frequent closed itemsets

In this paper, we propose an algorithm to partition both the search space and the database for the parallel mining of frequent closed itemsets in large databases. The partition-ing of the search space is based on splitting the power set lattice of the total item set to two sub-lattices. Conditional databases are used to partition the large database. The combination of the search space and database partitioning allows parallel processors to mine the frequent closed item-sets independently and thus minimizes the interprocessor communication and synchronization. The partitioning also ensures the load balance among the parallel processors