The Research And Application For Multidimensional Association Rules Algorithm Based On Multidimensional Predicate Index Tree

Mining association rules in data mining. as an important issue, has been widely used in major commercial areas which makes it become one of the most mature and important research in data mining.Multidimensional association rules as an important form of association rules, has also been rapidly developed into a research hotspot in recent years.First of all, this paper analyzes two main ways to study the current multi-dimensional association rule mining, one is to extend the association mining algorithms,so it can directly applied to multi-dimensional database and derived from multi-dimensional association rules, which. has good scalability and capable of handling large amounts of data, but it costs much about I/O for handling of the predicate-dimensional.So this method is less efficient in the whole. Another is to use the data cube technology for multi-dimensional association rule mining, this method has good I/O performance, especially when the data cube is small, you can use multi-dimensional arrays for the effective realization, but for a large database, construction and maintenance of the corresponding data cube is extremely expensive.Then,we propose an efficient algorithm for mining multidimensial association rules for the advantages and disadvantages of the two methods,which combine data cube technique with frequent itemsets mining technique efficiently by constructing a new data struct——MDPI-tree(Multi-dimensional predicate index tree), MDPI-tree is compose of DP-tree(Dimensioal predicate tree) structure and FP-tree structure. The algorithm adopted the idea of divide and rule, Firstly,we can build a data cube for the dimensional predicate to maximize compress storage space. Secondly, we can use the FP-Growth algorithm to mine the frequent itemsets constrained by dimensional predicate, Thus greatly reducing the time for information processing.The algorithm not only make use of the data cube which can effectively deal with multi-dimensional numerical measure, but also efficiently handle the affairs of the items of information which can explores both inter-dimension and hybrid-dimension association rules.Finally, This paper use vc++ 6.0, sq1 server 2000 as experimental platform and verify the performance of the MDPI algorithm by experiment.The experimental results show that the MDPI algorithm not only has good I/O performance, but also has good flexibility and stability. On the order hand, the MDPI algorithm is applied to cross-selling of a mobile communication company, and the results show that the multi-dimensional association rules mined by the new algorithm has some commercial value, which can provide decision analysis with basis for decision making.