Font Size: a A A

The Research And Implementation Of Association Rule Data Mining Algorithm

Posted on:2006-09-23Degree:MasterType:Thesis
Country:ChinaCandidate:Q L ZhaoFull Text:PDF
GTID:2168360152489837Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the development of information technology,Data Mining has been paid attention extensively,which stimulates more and more people who work in the field to research deeply into it.As we know,Data Mining has a large research scope.Association rules data mining is one of important research subject in it.Deeply researching into the subject has most important values not only on theoretics but also on applications.In the research of association rules data mining,the algorithms research is its key part for mining association rules.Many highly efficient algorithms in the field have been put forward for mining association rules so far. In the thesis,we analyse and summarize some useful algorithms presented by former researchers at first,and then we presented three new association rules algorithms for solving some relative problems in association rules data mining as follows: Firstly,we present an algorithm,MFP-Miner,for mining maximal frequent itemsets from transaction database.The algorithm needn't generate candidate itemsets and scan database, because it employs FP-Tree(frequent pattern tree) to store trasanctions in database compressedly and takes full advantage of the characteristic of FP-Tree.Consequently,the complexity of space and time the algorithm used is reduced obviously in the procedure of data mining.In a thorough experimental analysis of our algorithm on real and synthetic data,we isolate the effect of the individual components of the algorithm. Our performance numbers show that MFP-Miner outperforms previous work. Secondly,we present a new updating algorithm,UMFPA,for mining maximal frequent itemsets from transaction database when minimum support is changed by customer.The algorithm can efficiently mine maximal frequent itemsets with changed minimum support,because it makes the most of information provided by previous maximal frequent itemsets.From the experimental analysis of our algorithm we can conclude that the algorithm is highly efficient to the updating mining problems when the maximal frequent itemsets are very long. Finally,we present an integrated updating algorithm,IUMFIA,for mining maximal frequent itemsets when database and minimum support are changed simultaneously. The algorithm can efficiently mine new maximal frequent itemsets according to changed database and minimum support based on previous maximal frequent itemsets,because of using the characteristic of frequent items after database and minimum support are changed.From experimental analysis we can conclude that the algorithm is especially efficient to the application of integrated updating mining for maxial frequent itemsets.
Keywords/Search Tags:Data mining, Association rules, Maximal frequent itemset, Updating mining, Frequent pattern tree
PDF Full Text Request
Related items