Research And Improvement Of Multi-layer Association Rule Mining Algorithm Based On Clustering

In recent years,data mining has gradually got people's attention,and has evolved into a mainstream technology,so people are now more concerned with researching or predicting some patterns of people's behaviors by analyzing data.‘Shopping basket analysis' let more people put their attention to association rules,which is a classic function of data mining,and mostly to dig a single-layer association rules.Now people are not only interested in the rules of the same layer,they may also be interested in different layers of association rules.To dig out multi-layer association rules,we propose a method that combines association rules with clustering.At first,clustering the original transaction set and then having association rules analysis for clustered transaction sets,we can explore the rules of multi-layer association rules.In order to apply to multi-layer association rules mining of transaction data sets,this paper has made relevant improvements at every stage of the above.The research of this paper mainly focuses on the following four aspects:1.This paper summarizes the related theoretical knowledge about association rules,clustering concepts and multi-layer association rules mining in data mining,and analyzes the other academic achievements in recent years.At the same time,some related association rules and clustering algorithms are introduced.The association rules are mainly introduced Apriori and FP-Growth;The clustering mainly introduced K-Means and K-Mediods;2.Designing a specific measurement algorithm for SDS.In order to improve the quality of clustering,we improved SMC and Jaccard method to calculate the correlation coefficient matrix between the data objects in the data set of IBM.3.The improvement of association rule algorithm.On the basis of the FP-Growth algorithm,some improvements are put forward: It is necessary to reduce the size of the FP-Tree and save the memory by combining the nodes with the same number of support;Replacing the head table with a hash head table,which can speed up the search efficiency;4.Multi-layer association rules mining.By combining the clustering and association rules,the purpose of mining multi-layer association rules is achieved.In this paper,in order to mine the multi-layer and even the association rules between the layers,the original transaction data set is divided according to the clustering result.Unlike the other algorithms,the data set needs to merge the data objects in a cluster and cluster Name instead.This way,you can not only keep the information between different layers in the data set to mine multi-layer association rules,but also reduce the size of the data set.