A Model Driven Architecture Approach to Generate Multidimensional Schemas of Data Warehouses

Abstract

Over the past decade, the concept of data warehousing has been widely accepted. The main reason for building data warehouses is to improve the quality of information in order to achieve specific business objectives such as competitive advantage or improved decision-making. However, there is no formal method for deriving a multidimensional schema from heterogeneous databases that is recognized as a standard by the OMG and the professionals of the field. Which is why, in this paper, we present a model-driven approach (MDA) for the design of data warehouses. To apply the MDA approach to the Data warehouse construction process, we describe a multidimensional meta-model and specify a set of transformations from a UML meta-model which is mapped to a multidimensional meta-model. The transformation rules are programmed by the Query View Transformation (QVT) language. A case study illustrates our approach. It demonstrates how it reinforces the components traceability and reusability and how it globally improves the modeler’s efficiency. Furthermore, the use of the UML, as a technique to build data warehouses, is an important facilitator which prepares our further work to automate this approach.