Authors:
Redha Benhissen
;
Fadila Bentayeb
and
Omar Boussaid
Affiliation:
ERIC Laboratory, University of Lyon 2, 5 Av. Pierre Mendès, Bron, France
Keyword(s):
Temporal Data Warehouse, Multidimensional Model, Data Evolution, Slowly Changing Dimension, Graph-Based Database, Temporal Query, NoSQL.
Abstract:
Nowadays, companies are focusing on overhauling their data architecture, consolidating data and discarding legacy systems. Big data has a great impact on businesses since it helps companies to efficiently manage and analyse large volumes of data. In business intelligence and especially decision-making, data warehouses support OLAP technology, and they have been very useful for the efficient analysis of structured data. A data warehouse is built by collecting data from several data sources. However, big data refers to large sets of unstructured, semi-structured or structured data obtained from numerous sources. Many changes in the content and structure of these sources can occur. Therefore, these changes have to be reflected in the data warehouse using the bi-temporal approach for the data and versioning for the schema. In this paper, we propose a temporal multidimensional model using a graph formalism for multi-version data warehouses that is able to integrate the changes that occur
in the data sources. The approach is based on multi-version evolution for schema changes and the bi-temporal labelling of the entities, as well as the relationships between them, for data evolution. Our proposal provides flexibility to the evolution of a data warehouse by increasing the analysis possibilities for users with the decision support system, and it allows flexible temporal queries to provide consistent results. We will present the overall approach, with a focus on the evolutionary treatment of the data, including dimensional changes. We validate our approach with a case study that illustrates temporal queries, and we carry out runtime performance tests for graph data warehouses.
(More)