A Novel Prefix Graph Based Closed Frequent Itemsets Mining Algorithm

The key points of the frequent closed itemsets mining are based on two main steps: search space browsing and closed itemsets detecting. This paper presents NPG_mining, a novel prefix graph based algorithm for mining closed frequent itemsets. The new approach has constructed an efficient prefix graph structure and use variable length bit vectors to present the relationship between the database and its items. Based on die closure equivalence concept, the algorithm has created a efficient closed core item generating technology, which can identify the possibility of a prefix itemset turning into a closed frequent itemset without keeping the existing closed frequent itemsets in the main memory. Our performance study shows that the pruning efficience and scalability using NPG_mingning is better than PGMiner.