Designing data cubes in OLAP systems: a decision makers' requirements-based approach

摘要

Business Intelligence systems rely on an integrated, consistent, and certified information repository called the Data Warehouse (DW) that is periodically fed with operational data. In the decision-making process, the analyzed data are usually stored in the DW in the form of multidimensional cubes. These cubes are queried interactively by the decision makers, according to the online analytical processing paradigm. In larger companies with multiple subsidiaries, the frequent expression of new business needs requires the creation of new data cubes which generate a large number of cubes to be manipulated. The inevitable complexity and heterogeneity of data cubes make it difficult to design data cubes. The decision maker can precisely express his needs through a query in natural language which consists of a set of analysis indicators (measures, dimensions) separated by the AND operator. However, the decision maker's need may be incomplete. Indeed, he usually has a cube that represents part of his needs and he may want to complete it or enrich it with other cubes that are unknown to him. To deal with these situations, we propose in this paper an approach that addresses the problem of designing and constructing data cubes where the expressed need is scattered over more than one cube. Our goal is to enable decision makers to analyze all of their needs using just one cube. Our approach consists of two variants: a variant that is based on analysis indicators, and another based on the known cube. We present the validation of our approach by means of a tool, called "Design-Cubes-Query" that implements our approach and we show its use through a case study.