Definition and Usage
DW in addition to OLAP technologies intend to be an innovative decision support for business intelligence and knowledge discovery. It has now become a leading topic in business organizations as well as in the research community. The main motivation is to take benefits from the enormous amount of data in distributed and heterogeneous databases to enhance data analysis and decision making (Kimball & Ross, 2002).
A DW is a subject-oriented, integrated, nonvolatile, and time-variant collection of data stored in a single site repository and collected from multiple sources (Inmon, 1996). Information in the DW is organized following a multidimensional model in order to allow precomputation and fast access to summarized data in support of management’s decisions. This multidimensional model organizes data in analysis axes called dimensions. Analyzed subjects or facts are characterized by metrics called measures. Dimensions can be organized following hierarchy schemas, thus allowing navigation through different levels of detail of analysis.
An OLAP server calculates and optimizes the hypercube, that is, the set of fact values for all combinations of dimension instances (called members). In order to optimize accesses to the data, query results are precalculated in the form of aggregates. This allows the decision makers to explore the different dimensions at different granularities. This analysis process is conducted by navigating into the multidimensional cube through some OLAP operators (roll-up, drill-down, slice, rotate, etc.).
Finally, interactive user interfaces (OLAP clients) have been developed to support knowledge discovery, promoting the iterative nature of the analysis process. OLAP clients visually represent the multidimensional structure of the hypercube and formulize multidimensional queries. The most adopted data presentation paradigm is the pivot table, a 2-D spreadsheet with associated subtotals and totals. It supports complex data by nesting several dimensions on the x- or y-axis and displaying data on multiple pages.
The three-tier architecture composed of a DW, an OLAP server, and an OLAP client effectively allows multidimensional analysis.
In the following, we will use the conceptual multidimensional model MultiDimEr (Malinowski & Zimányi, 2006) in order to describe our proposal. The details of the model are presented in Figure 1.
Figure 1. MultiDimEr conceptual multidimensional model