Research On Algorithms For Association Rules Mining Based On FP-Tree

Data mining technology is one of the frontier research fields of the information science. Mining association rules of dataset is one of the most important research fields in data mining, which can reflect the relationships among items and is widely used in marketing, medicine, telecommunications, agriculture and many other fields. This paper focuses on the mining algorithms of association rules based on FP-Tree.In order to better analyze the characteristics of different databases and the efficiency of different algorithms in database, the concept of association saturation degree is defined, and the method based on association saturation to optimize the typical FP-growth is also proposed in the paper.FP-growth, which is based on FP-Tree, is unlike Apriori which creates a lot of candidates, but it still needs a lot of traversal and recursive operations in building FP-Tree and mining association rules, which reduces the efficiency greatly. To solve this problem, a new structure of FP-Tree and a new algorithm DFP-growth which is based on DFP-Tree are proposed in this paper. The DFP-growth adds new items to save the address of successor node and predecessor node, which make the operation easier. LDFP-growth, which divides the large database into sub databases by the set of frequent 1-itemsets and mines the sub databases one by one, is proposed in this paper.The simulation results indicate that DFP-growth algorithm has a good performance in mining small database, while LDFP-growth has a good performance in mining large database, and performs better than FP-growth and DFP-growth when candidates grow to a large scale.The advantages and disadvantages of three proposed optimization methods are summarized at the end of the paper, and the steps of association rules mining based on DFP-growth and LDFP-growth are also presented.