PrePost+: An efficient N-lists-based algorithm for mining frequent itemsets via Children-Parent Equivalence pruning

N-list is a novel data structure proposed in recent years. It has been proven to be very efficient for mining frequent itemsets. In this paper, we present PrePost + , a high-performance algorithm for mining frequent itemsets. It employs N-list to represent itemsets and directly discovers frequent itemsets using a set-enumeration search tree. Especially, it employs an efficient pruning strategy named Children–Parent Equivalence pruning to greatly reduce the search space. We have conducted extensive experiments to evaluate PrePost + against three state-of-the-art algorithms, which are PrePost, FIN, and FP-growth ⁄ , on six various real datasets. The experimental results show that PrePost + is always the fastest one on all datasets. Moreover, PrePost + also demonstrates good performance in terms of memory consumption since it use only a litter more memory than FP-growth ⁄ and less memory than PrePost and FIN. Frequent itemset mining, first proposed by Agrawal, Imielinski, and Swami (1993), has become a popular data mining technique and plays an fundamental role in many important data mining tasks such as mining associations, correlations, episodes, and etc. Since the first proposal of this new data mining task, there have been hundreds of follow-up research publications (Han, Cheng, Xin, & Yan, 2007). Although lots of algorithms have been proposed, how to design efficient mining methods is still one of several key research problems yet to be solved In recent years, we present an algorithm called PrePost (Deng et al., 2012) for mining frequent itemsets. The high efficiency of PrePost is achieved by: (1) employing a novel data structure named N-list to represent itemsets; and (2) adopting single path property of N-list to directly discovery frequent itemsets without generating candidate itemsets in some cases. The experiments in Deng et al. (2012) show that PrePost runs fast than some state-of-the-art mining algorithms including FP-growth and FP-growth ⁄. Although PrePost adopts single path property of N-list to prune the search space, it still incurs the problem of too many candidates because it employs Apriori-like approach for mining frequent item-sets. In this paper, we propose a new algorithm called PrePost + , which can effectively avoid the above problem. PrePost + employs N-list to represent itemsets and directly discovers frequent item-sets in an itemset & N-list search tree. For avoiding repetitive search, it also adopts Children–Parent Equivalence pruning to greatly reduce the search space. For evaluating the performance of PrePost + , we conduct …