Research On Utility-oriented Pattern Mining

With the rapid development of the Internet,the information exchanges become convenient and the data quantity is growing exceedingly,which is of great value.How to mine data has a more broaden influence on every industry.Association rules are the core to the data mining while association rules only count the times of a pattern presented;as a consequent,some lower-presented but high utility patterns can not be mined out.The high utility pattern mining algorithmhas solved this problem successfully.In recent years,the efficient pattern mining has become a key research direction in data mining.There are many types of patterns,from the initial Boolean-based association rules,the model has been widely expanded,the model with the time attribute to produce the form of the sequence pattern.This article is focused on high utiliy sequential pattern mining algorithm based on particle swarm optimization,on the terms of high utility pattern mining algorithm as well as the sequential data set.The main results were summarized as follows:In the case of traditional dataset with utility information,this paper proposes a high utility pattern mining algorithm based on particle swarm optimization HUPMSUPT-PSO,which designs the SUP-Tree structure.This structure saves the upper bound on the model,which plays an important role in the pruning of the model,and pre-evaluates the particles before evaluating the particle phase,which greatly improves the efficiency of the algorithm.Experimental results show that the HUPMSUPT-PSO is more efficient than the current algorithm HUIM-BPSO+ for high utility mining based on particle swarm optimization.Transactions in database in the form of sequence produces high utility sequential pattern mining algorithm.The search space of sequence pattern is bigger than that of normal pattern.So its computational complexity is much bigger.The study of high utility sequential pattern mining algorithm is relatively rare.But not considering the circumstance of the external utility in sequential database is negative.Faced with external utility which includes negative items,the High Utility Sequential pattern mining algorithm EHUSN with negative items is put forwardat the first place.It proposes 1-2-UM and 2-2-UM these two structural models on the base of FHM algorithm,and the model can prune noncandidate sequence when combined with utility information list,therefore,data mining algo-rithm is becoming more efficient.