Research And Implementation Of Multidimensional Data Modeling

With the development of information technology,people have accumulated a large amount of historical data through online transaction processing(OLTP).They are also increasingly aware of these historical data which contains information on the importance of decision support.The results show that the traditional data model,such as Entity-Relationship(E-R) model,can't accurately convey the data structure and semantics of the data warehouse. It's difficult to effectively support online analytical processing(OLAP),so people consider the establishment of a multidimensional data model,and organize and manage these data through the data warehouse and OLAP.Afer referencing to some multidimensional data modeling tools and multidimensional data model,inplemented a comprehensive multi-dimensional data modeling tool,it can help companines better manage historical data.The tool has three major modules:module of multi-dimensional conceptual model,module of multi-dimensional physical model and module of reverse engineering.Reverse engineering consists of two parts:OLAP to the multi-dimensional conceptual model and the relational database to the multi-dimensional physical model.To create multi-dimensional conceptual,it absorb the existing popular design,for example the whole conceptual model contains dimension including personal dimension,sharing dimension and collapse dimension,measure and fact and so on.The fact tuple consists of primarykey of dimension and measure.moreover,the tool increases the sub-dimension design.Taking into account some of the time dimension of the property changes,these properties can be extracted fom a separate sub-dimension,so changes in sub-dimensino will not affect the parent dimension.It can help maintain model and control the excessive growth of the dimension table.Or it may exist between the different dimensions of the same dimension attribute,then we can extract the common properties of these to form a separate sub-dimension.Dimension of being selected can absorb the way by reference to the sub-dimension.The former case snowing the dimension,equivalent to the star model is transformed into snow model.While the later case propose the formation of different dimensions of the separate dimension of the same property,it equals to transform the star model into stellar population model.Physical model transformed by multi-dimensional conceptual model is the same with entity-ralationship model.Multi-dimensional physica model includes two table type:dimension table and fact tabe.They are respectively transfromed from the dimension and fact.Fact table foreign key way through the dimension table is connected to the primary key.The tool also can generate SQL script,then it can change the multi-dimensional physical model into relational tables.Taking into account the XML view can be reflected from the model structure,that is its good scalability and versatility,multi-dimensional physical model and the referred multi-dimensional conceptual model in the last paragraph are imported and exported by it.Resulting XML document is the multi-dimensional data model metadata.In the OLAP to multi-dimensional conceptual reverse engineering,it extracts from multidimensional data model under the interface provided by DM company,convert it to the conceptual model in the tool,modifying the model ,then input the refined model into the OLAP server,so it achieve the optimal of OLAP multidimensional data model in the circle.Morover we will find that the existing multidimensional data model is the same with the multidimensional conceptual to be constructed.Similar to the above case,in entity-relationship database to multidimensional physical model reversing engineering.We extracted the relation tables in DM database through the JDBC interface provided by DM database,transforming them into multidimensional physical model in the modeling tool,it also achieved optimization of entity-relationgship model.