BitApriori: An Apriori-Based Frequent Itemsets Mining Using Bit Streams

Generating, pruning and counting itemset candidates are important steps in Apriori frequent itemset mining. Unfortunately, their computation time are too expensive. In this paper, we propose a new method using Bit Stream to improve their speed. At the begining, the 1-itemsets are found out and sorted according to the decline of count. By that way, a map of all attributes would be created. After that, each attribute will be presented by 1 bit. At last, the generating and pruning itemset candidates are processed by LOGIC operations which are not cost much of computation time. For experiments we compare our method with some Apriori-based state of the arts.