Research On Mining Algorithm Of Association Rules Based On Frequent Pattern Tree

With the rapid development of human society,database technology and the technology of information management has been widely popularized.And the amount of data accumulated by human society has grown at an exponential rate.In the face of vast data,how to efficiently and quickly obtain "valuable" information from the data to serve human beings is an important problem that we are facing.Therefore,data mining emerges.Data mining is the process of acquiring real valuable and undiscovere rules from data.And the association rule is one of the important research directions of data mining.It was proposed in 1993 by Agrawal.It's main purpose is to discover the relationship between different sets of items in massive data.The association rule reveals the relationship between different itemsets,which is widely used in finance,medicine,shopping malls,telecommunications,industry and so on.Since the introduction of the association rule,There are two classic association rules algorithms have emerged:the Apriori algorithm and the FP-growth algorithm.As the most widely used FP-growth algorithm,it does not need to generate candidate sets during the mining process.However,when the data set is so huge and the RAM is limited,the traditional FP-growth algorithm will meet the limitations such as slow running speed and large memory usage.In order to improve the shortcomings of the traditional FP-growth algorithm,this paper proposes a DOFP-Mine(decomposed ordered frequent pattern tree mining)algorithm for large-scale data mining based on the traditional FP-growth algorithm.The DOFPmine algorithm firstly overcomes the problem that the traditional frequent pattern tree mining algorithm can not be effectively applied to large-scale data mining and decomposes the data set into data subsets for mining,.Then the algorithm uses ordered construction and two-dimensional array-assisted mining.The method optimizes the efficiency of the algorithm in terms of time and space.Experiments show that the DOFP-Mine algorithm outperforms the traditional frequent pattern tree algorithm in large-scale data mining.