Research And Application Of Hybrid-driven Data Warehouse Modeling Method

Multi-dimensional modeling approach is one of main ways to build data warehouse. The current multi-dimensional modeling methods include: data-driven, demand-driven and goal-driven, etc., but these methods are difficult to take into account both the good show performance on multi-dimensional model and high usability of the data warehouse. Therefore it is helpful to propose an effective multi-dimensional model which can meet the requirements of data application and data presenting.In this paper we proposed a hybrid multi-dimensional modeling method which can generate the candidate set of facts and dimensions based on ontology, analyze the business requirement and business processes to get specific fact and dimensions and generate multi-dimensional model. A prototype system is proposed to achieve a hybrid modeling method and verify the availability of the method.The main research work as following:First, the framework of hybrid modeling method is proposed through analyzing the status of multi-dimensional modeling approaches and the effect of them in the practical application. The framework include: based on the input ontology the potential facts and dimensions are defined according to the ontological constraints and the candidate set of fact and dimensions are generated; from a business point of view, the logic model composed of business concepts is generated through the analysis of the specific business needs and business processes and then the business concepts and should be associated with ontology concepts to get multi-dimensional model.Secondly, through the analysis of the relationship between fact and dimension, the many-to-many correspondence between fact and dimension is gotten and formal method is used to express this relationship. And the constraints of the ontology properties are used to describe the relationship between fact and dimension. Thus through the analysis of the ontology concepts and attributes, the constraints of the ontology properties are used to distinguish the fact-dimension relationship and the candidate set of facts and dimensions can be generated.Thirdly, for different business requirements, the logic model composed of business concepts is generated through the analysis of business use cases and processes between these use cases. Then the business concepts and should be associated with concepts in candidate set of facts and dimensions to get multi-dimensional model composed of ontology concepts for the special business requirements.Finally, the prototype system is designed and the main system framework includes generating candidate set of facts and dimensions module and the establishment of multi-dimensional model module. In order to verify feasibility of the hybrid multi-dimensional modeling method, the whole process of the method is achieved combining business needs and data on medical fields. And the data mining to the established multi-dimensional model are used as a case study to verify availability. Compared with data-driven and demand-driven method, hybrid-driven method has characteristics on data presenting and data application and has advantage in modeling risk, time, user participation, and other areas.