Algorithm For Mining Association Rules Based On Clustering

At present, a lot of researches on the association rule mining is mainly focused on frequent itemsets mining association rules cutting direction, frequent itemsets mining is one of the most important researches, there are a lot of excellent algorithms on it, such as Apriori, FP-growth and so on. FP-growth algorithm needs to be carried out on the same path repeatedly backtracking, thus affecting the efficiency of mining. Clustering the data mining is an important research area, for other operating data preprocessing, mining association rules based on clustering has a very good sense. Algorithm for mining association rules based on clustering, Though there are not so many researches on it, it still has achieved some results. In this paper, the association rules are based on clustering research.The content of this article is divided into two aspects:one is the FP-growth algorithm mining conditional pattern base repeated backtracking problem; another one is the problem of mining association rules based on clustering.To study repeated issues. This paper first expounds two classic frequent itemsets such as mining algorithm Apriori and FP-growth, then analysises of the advantages and disadvantages of the FP-growth algorithm. The advantage is that FP-growth algorithm is a candidate set of algorithms With frequent pattern tree FP-tree, cleverly compressed data structure, eliminating the need for reliance on the candidate set and greatly reduce the pressure on the I/O; While the disadvantage is that it’s too complex, because of FP-tree structure whose mode based in mining conditions needs repeated on the same path backtracking. Based on these two shortcomings, I propose a FP-tree transformation method, so that it needn’t to record the address of each node but to remember the address of the leaf node of each path, What’s more, it can further propose a new non-repeat algorithm named FFP-growth algorithm. The algorithm in mining conditional pattern base, just backtracking time on the same path can be obtained conditional pattern base of all nodes on the entire path experiments prove that this method improves the efficiency of mining.To study mining association rules based on clustering, this paper first expounds the basic knowledge of clustering and mining association rules.Then proposes FP-growth algorithm-CFP-growth algorithm which based on clustering.The algorithm is:firstly, gather data and put them into clustering table, and then built frequent pattern subtree by the clustering table. Finally, adopt the FP-growth algorithm to mining frequent pattern from the sub-tree of the frequent pattern. Because each of the single frequent pattern subtree amount of data is smaller than the whole frequent pattern tree, so it is relatively easier to mining frequent pattern. The experiment is done in FoodMart database, and the results show it has a good effect on mining frequent patterns.