New algorithm for mining frequent itemsets in sparse database

This paper presents novel algorithm for mining frequent itemsets in sparse database, compared with existing algorithm our algorithm has visible advantage. With this algorithm, the scans is less in transaction database, only one time in little and middle transaction database, and not more than two times in large database. In the algorithm, when the transaction database is scanned, the transaction items are saved in unit triplet, and the count of every transaction item is saved in 1-dimension array so that the frequent itemsets are generated in memory. So I/O spending is reduced greatly. The experimental results show that our algorithm is very promising.