Multidimensional Association Rules Mining In Vertically Partitioned Data

Mining distributed association rules in distributed data is a very important researchdirection. By studying the existing paper we found that the current research is focused onthe association rule mining of the data which is horizontally partitioned, and few ofresearch focus on the vertically partitioned data. The existing algorithms are limited tosingle dimension and the bool-type data. There is no algorithm deal with the non-bool andmultidimensional data. After the analysis of the home and abroad research status, thispaper focus on mining association rules in vertically partitioned multidimensional data.Firstly, the existing algorithms are limited to single dimension and the bool-type data.There is no algorithm deal with the non-bool and multidimensional data, so a newalgrithem was proposed to mine in two parts vertically partitioned non-bool data namedRTVP.Secondly, in the background of subject, there are two kinds of relationship:1)the datacenter and one applicant;2) the data center and many applicants. In order to minemultidimensional association rules in the condition of the two kinds of relationship, basedon the principle of RTVP, MDOA and MDMA were proposed, association with thetechnology of data cube.Finally, a experimental scheme was designed, chose the dataset for the experiment. Inorder to simulate the two kinds of relationships of data center and applicant, the datasetwas split and disposed, and the corresponding data cubes were created. Then using VisualStudio2008and Analysis Services develop the two algrithem paragrams, and verify thealgrithems of MDOA and MDMA.