FUP Algorithm Optimization Based On Compression Matrix

Association rule mining has always been an important research field in data mining.For the traditional association rule mining,scholars have proposed many classical algorithms,as well as the optimization ideas of these classical algorithms.Apriori algorithm based on iteration of candidate set and FP-Growth algorithm based on tree structure mining are typical representatives.According to literature research,most of the optimization schemes of mining algorithms at this stage are based on these two algorithms.However,when the data set changes,the existing mining results will become unreliable.Such algorithms need to abandon the existing mining results,and then re-mine the entire data set.This paper calls this mining method static association rule mining.With the arrival of the era of big data,the data set is getting larger and larger,and the efficiency of this static mining algorithm is very low,especially when the data set changes relatively small.We found that traditional static association rules mining cannot solve most of the problems in the real world.Therefore,incremental association rules mining has been widely studied by scholars.Thus far,many incremental association rules mining algorithms have been proposed by scholars.The Fast Update Pruning(FUP)algorithm is a classical way to solve association rules mining in the case of a database increase.The FUP algorithm reduces the number of computations by pruning the existing frequency itemset.However,the algorithm requires frequent transactional database scans,and this operation creates a performance bottleneck when the amount of data is large.To solve this problem,this paper combines the FUP algorithm with the compressed Boolean matrix based on the study of matrix compression and proposes a new incremental association rule mining algorithm named the FUP algorithm based on compression matrix(FBCM).This algorithm simply needs to scan the respective transactional and incremental databases one time.It establishes two compressible Boolean matrices and then performs association rule mining on the Boolean matrix.The algorithm reduces overhead,memory costs and computational burden through the compression matrix.Meanwhile,it effectively improves the computational efficiency of incremental association rules mining.