An Optimization to CHARM Algorithm for Mining Frequent Closed Itemsets

Frequent itemsets which are quite useful in many applications always suffer from their huge number and information redundancy. Frequent closed itemsets that provide a minimal and lossless presentation of all frequent itemsets are a solution to the problem. In past years, frequent closed itemsets mining (FCIM) has been extensively studied and many effective FCIM algorithms have been proposed. CHARM as one of the most famous FCIM algorithms is easily extended and applied due to its understandable and commonly used structure. However, CHARM faces a memory-inefficient challenge since it needs to maintain all closed itemsets in the memory to check if an itemset is closed or not. In this paper, we introduce a new method of identifying nonclosed itemsets with strict proofs and propose algorithm NEWCHARM, an optimization to CHARM for mining frequent closed itemsets that is more memory efficient. Moreover, we also prune the nonclosed itemsets timely in the search process so that the search space is reduced and with some optimized operations adopted, our algorithm also runs faster. Finally, experiments are conducted to compare NEWCHARM against CHARM on six datasets from two aspects in memory usage and execution time and the results show that our algorithm outperforms CHARM.