Study On Association Rules Mining Algorithm Based On FP-tree

Data mining, which is called knowledge discovery in database, has been attached importance by the field of international artificial intelligence and database. It is a process of discovering latent and interesting knowledge from plentiful data. As one of the important contents in data mining, association rule mining aims to discover the interesting connection or the correlation midst a set of objects in a database. It is applied widely in many fields.Apriori algorithm and FP-growth algorithm are two main algorithms in mining association rules which are based on frequent itremsets. Different from the former candidate set generation-and-test patterns of the Apriori-like algorithm, FP-growth algorithm takes the way of pattern fragment growth to mine association rule. It completes the Apriori algorithm and has good effect. But the FP-growth algorithm is not perfect. First, its performance seriously relies on the scale of the database. Next, it may generate and release millions of conditional pattern trees in the process of mining, and so on.This paper has completed the following research work to solve the problems of the Apriori algorithm and the FP-growth algorithm.(1) After knowing the actuality of the association rule mining, this paper mainly studies the FP-growth algorithm which is based on the FP-tree. It also analyzes and discusses the main existence problems.(2) This paper proposes a rear-inserting construction of frequent pattern tree based on projection technology. It constructs the FP-tree level by level based on projection technology and makes the best of the projection operation ability of the large-scale database. It solves the problem effectively which exist in the traditional FP-tree construction.(3) This paper has studied the implementation of the construction of the FP-tree and PRIFP-tree. It compares these two construction methods and analyzes the performance of them through experiment. The result shows that the PRIFP-tree construction is more efficient and scalable than traditional FP-tree construction, especially when the transactions are huge.Our research work is a practical improvement to the association rule mining algorithm. It has valuable reference to the research of association rule mining algorithm base on SQL.