Research On Frequent Pattern Mining Of Multi-mimimum Support And Parallel Computing Based On Similer Items

With the progress and development of the information society,information explosion has become the opportunity to face the challenges and challenges.Therefore,how to get useful knowledge in these information has become a hot topic.Data mining techniques are widely used to process large amounts of data stored in a database to extract the required information.Among them,frequent pattern mining is one of the most effective data mining techniques.However,the traditional frequent pattern mining algorithm has some drawbacks.In this paper,two improved algorithms are proposed for these problems.The main research contents are as follows:1.Introduce the concept of data mining,its preprocessing method,and the current situation of research at home and abroad,and then introduce the concept of association rules mining and association rules mining classic algorithm.2.In order to solve the problem that the frequent pattern mining algorithm takes a lot of time and memory in the mining process,we first use the structure of the enumerated enumerations and use the multi-minimum support to reduce the time and memory consumption.However,when the number of transactions in the transaction database is particularly large,it is obviously unscientific to define a unique minimum support threshold for each item.For the above problem,we first propose to classify the same items for the database transaction,and then assign a unique minimum support threshold to each class to be divided,and finally use the concept of backward closed attribute ordering and LCMS(minimum support level)to trim the search space,thus giving the FP-CME algorithm.The algorithm does not need to generate conditional candidate tree in the mining process,and can find the frequent pattern from the set enumeration tree directly.The simulation results show that the algorithm has a better performance than the traditional algorithm in the execution time and memory usage.3.FP-growth is one of the most classic algorithms in frequent pattern mining.However,the FP-growth algorithm is also problematic: when it is excavated in a large-scale data environment,the FP tree is too complex,The result is the difficulty of building FP-trees and the inefficient excavation.In view of the above problems,we propose a POFP-growth algorithm based on multi-minimum support.The algorithm is divided into two steps,the first step,the original database transaction items areclassified,divided into multiple classes and give the class label and the only support threshold.Then,the database data is shared prefix prefix,then the transaction database level is divided into N parts,and then build the local database FP-tree,and finally use the internal level join method to combine all the local FP-tree to get a complete FP-tree.In the second step,we use the item merging strategy to mine the frequent itemsets in the complete FP-tree under the constraint of the multi-minimum support threshold.Through the simulation experiments we can see that the proposed algorithm has a certain improvement in the execution time.