Author(s)

Y. Mai, J. Li & H. L. Viktor

Abstract

Dimensional modeling is a common modeling technique in data warehousing. It reflects a simple logical view of a data warehouse system. It can be easily mapped to a physical design. Traditional dimensionalmodeling is data-oriented and semantically informal. From a software engineering perspective, the informal notations and data-oriented feature are insufficient to tackle the complexity of large data warehouse projects. UML, with its well-defined semantics, is now a standard modeling language that is used to model the entire life cycle of a software system. UML has rich and extensible semantics. The combination of the knowledge in standard object-oriented modeling and dimensional modeling add variable semantics to dimensional modeling without losing its understandability. This paper proposes a metamodel for data warehouse dimensional modeling using UML. Based on this metamodel, we illustrate how to model the business process and data marts of a large mobile telephone company. Keywords: data warehouse, dimensional modeling, UML, metamodel. 1 Introduction and related work In recent years, data warehouses have gained increasing popularity and are becoming a business growth strategy. A data warehouse is essentially a data container, which contains complete and historical business data from numerous operational sources. The data, as contained in a data warehouse, are used to analyze business, help predicting the organizational growth and improve customer relationships. In essence, a data warehouse is a queryable data source that exists to answer questions people have about the organization. These queries thus reflect the way that managers think about their organization and assist them to make sense of the data, form policies and to make informed decisions.

Keywords

data warehouse, dimensional modeling, UML, metamodel.