The Research Of Heterogeneous Entity Association Method Based On Moss

In MOSS 2007, only the entities in the same business application can associate with each other. Business data catalog can integrate many kinds of business applications from back-end server applications. The entities from different business applications are called Heterogeneous Entity. There maybe associations exist among heterogeneous entities. It's difficult for business analyst to find associations among heterogeneous entities. Therefore, how to mine the associations quickly and achieve the associations in MOSS is very important.Firstly, in order to determine the similarity of words, analyze the structural characteristics of Business Data Catalog in MOSS, and give the criterion function. On this basis, Extended Name Vector is introducted, and propose ENV (Extended Name Vector) algorithm according to Business Data Catalog. ENV algorithm based on Chinese Synonym Dictionary and WordNet, generates Extended Name Vector set that contains English synonyms, Chinese synonyms and their synonyms, combinations or abbreviations according to entity of Busi- ness Data Catalog. ENV provides a theoretical basis for the MHEA algorithm.Secondly, in order to mine the associations among heterogeneous entities of MOSS, we propose MHEA algorithm based on Extended Name Vectors. According to the entity structure of business applications in MOSS, we give the method that how to determine whether there are associations between two heterogeneous entities, and mine the associations among heterogeneous entities by the attributes of their FindSpecific methods on this basis. MHEA algorithm outputs heterogeneous entities association matrices which provide a theoretical basis for the business analysts to develop solutions.Finally, using the BDC(Business data catalog) of AGC Automotive China Co.,Ltd Management Information System to test the efficiency of the method proposed in this paper. In the experiments of this paper, the model of heterogeneous entity association is proposed. Then we achieve the associations among heterogeneous entities by the model, and analyze the performance of the method according to the results. The feasible of our proposal have been demonstrated through the experiments.