Mining N-most interesting closed itemsets

This paper presents an alternative mining task, called N-most interesting closed itemset mining, for generating the desired number of the most frequent closed itemsets in different lengths. The N-most interesting closed itemset mining is proposed to avoid generating redundant itemsets, and difficultly giving a minimum support. An efficient algorithm, called NCLOSED, is developed for mining N-most interesting closed itemsets. The algorithm directly generates closed itemsets without keeping candidates in memory. Moreover, the NCLOSED algorithm reduces search space by detecting and discarding duplicated closed itemsets. In addition, closed itemsets are discovered in a descending order of their support values.