A novel process-based association rule approach through maximal frequent itemsets for big data processing

Abstract

The maximal frequent itemsets issue in big data processing has become a hot research topic. Most of the previous work on big data processing directly analyzes the data through the existing approaches, which would cause problems of redundant computation, high time complexity, and large storage space. To solve the problems, this paper proposes a Heuristic MapReduce-based Association rule approach through Maximal frequent itemsets mining, HMAM. The main idea is: At first, by directly operating on the transaction database, we allocate transactions to different processing nodes and group all transactions according to dimension. Then, we screen the most frequent transactions from each transaction set using the Bitmap-Sort and obtain best-transaction-set through aggregating all the transaction-elects of each transaction set. The current candidate maximal frequent itemsets can be acquired by removing sub-transactions in terms of the inclusion relations of the items in best-transaction-set. At the same time, each subset of sub-transactions in the candidate maximal frequent itemsets is discarded from all transaction sets. Then the final candidate maximal frequent itemsets can be obtained by iteration until each transaction set is empty. Finally, we achieve the acquisition of maximal frequent itemsets by employing the minimum support threshold. The experimental results demonstrate that compared with the existing approaches, HMAM significantly avoids producing a large number of candidate itmesets resulting from join operation, accelerates the speed of mining the maximal frequent itemsets, and improves the utilization rate of resources simultaneously.