Research And Implement Of A Matrix Based Paralleled Frequent Itemset Mining Algorithm

Frequent itemsets mining has become an important research direction in data mining field. And it has a widely used in the correlation analysis, and intrusion detection fields. Researchers also proposed a lot of efficient frequent itemsets mining algorithm. But with the growth of the quantity of the data, frequent itemsets mining algorithm in serial mode cannot do miner job quickly. So proposed a efficient paralleled frequent itemsets mining algorithm is become more and more important for researchers and it also become a key point of the data mining field. This proposed a new paralleled frequent itemsets mining algorithm based matrix (we may called MPHP-Miner) after researching the existing paralleled frequent itemsets mining algorithm.Firstly, we are analyzing the development status of frequent itemsets. Then described a several classic frequent itemset mining algorithm, and also discusses the advantages and disadvantages of each algorithm.Secondly, we are researching and analyzing several existing paralleled frequent itemsets mining algorithm. Analyze the data structures and paralleled strategy and balancing strategy of the algorithm. And also discusses the advantages and disadvantages of each algorithm.Thirdly, MPHP-Miner we proposed in this paper using matrix to storage itemsets to reduce memory spending. This algorithm also using paralleled pattern calculates frequent itemsets to reduce time consuming.Forthtly, Realized our MPHP-Miner algorithm on parallel using X10.Through the experiment we proved MPHP-Miner is effective and feasible. And it is able to reduce the memory usage and operation time effectively. So MPHP-Miner is a good paralleled frequent itemsets mining algorithm.