Algorithms For Fuzzy And Objective-Oriented Association Rules Mining Based On FP-tree

Association rules mining is a very important problem in data mining. Traditional association rules mining is called boolean associaion rules mining.People have done abundant work for them,and have obtained achievements.Fuzzy and objective-oriented association rules mining are hot issues.The study on their algorithms has been acquired progress.But Most algorithms for them are based on Apriori or Apriori-like, they exist the problems:repeatedly scan the database and check a large set of candidates by pattern matching.In order to improve mining efficiency,we study algorithm for Fuzzy and objective-oriented association rules mining based on FP-tree. The main contributions of this paper are as follows:We present an efficient algorithm for mining fuzzy frequent itemsets,called FMF.We use FFP tree structure to store frequent item sets imformation,and store IDs of transactions related with fuzzy item in tree nodes.In FMF,we can count a fuzzy itemsets support through finding all trasactions including them.We needn't to scan database all.To generate itemset {a}+X(i.e. super set of iemset X) according to constrained subtree of itemset X,if item"a"isn't a fuzzy item,we don't scan database in addition.It can be generated by FFP tree.We propose two order methods for constructing FFP tree.One is that sorting database attributes holding frequent item in ascending order of their nodes number in FFP tree.Another is that sorting frequent item of not fuzzy attributes in descending order of their support firstly,then sorting database fuzzy attributes with frequent item in ascending order of their nodes number in FFP tree.Our experimental results show that although FMF needs more space costly than the algorithms based on Apriori,its time costly is obviously lower than the latter. The second order method of FMF is more efficient than the first one,and the advantage will be more obvious with support limit lower.The OOA mining hopes to mine all association rules based on a given objective, support, confidence and utility. We present a new approach to mine OOA rules using FP-growth algorithm. Our experimental results show that the algorithm is more...