Ontology-Based Big Dimension Modeling in Data Warehouse Schema Design

Xiufeng Liu, Nadeem Iftikhar

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review


During data warehouse schema design, designers often encounter how to model big dimensions that typically contain a large number of attributes and records. To investigate effective approaches for modeling big dimensions is necessary in order to achieve better query performance, with respect to response time. In most cases, the big dimension modeling process is complicated since it usually requires accurate description of business semantics, multiple design revisions and comprehensive testings. In this paper, we present the design methods for modeling big dimensions, which include horizontal partitioning, vertical partitioning and their hybrid. We formalize the design methods, and propose an algorithm that describes the modeling process from an OWL ontology to a data warehouse schema. In addition, this paper also presents an effective ontology-based tool to automate the modeling process. The tool can automatically generate the data warehouse schema from the ontology of describing the terms and business semantics for the big dimension. In case of any change in the requirements, we only need to modify the ontology, and re-generate the schema using the tool. This paper also evaluates the proposed methods based on sample sales data mart.
Original languageEnglish
Title of host publicationBusiness Information Systems. Proceedings
EditorsWitold Abramowicz
Publication date2013
ISBN (Print)978-3-642-38365-6
ISBN (Electronic)978-3-642-38366-3
Publication statusPublished - 2013
Externally publishedYes
Event16th International Conference on Business Information Systems - Poznan, Poland
Duration: 19 Jun 201321 Jun 2013
Conference number: 16


Conference16th International Conference on Business Information Systems
SeriesLecture Notes in Business Information Processing


Dive into the research topics of 'Ontology-Based Big Dimension Modeling in Data Warehouse Schema Design'. Together they form a unique fingerprint.

Cite this