An efficient procedure for mining statistically significant frequent itemsets

Abstract

We suggest the original procedure for frequent itemsets generation, which is more efficient than the appropriate procedure of the well known Apriori algorithm. The correctness of the procedure is based on a special structure called Rymon tree. For its implementation, we suggest a modified sort-merge-join algorithm. Finally, we explain how the support measure, which is used in Apriori algorithm, gives statistically significant frequent itemsets.