Temporale Auswertungsformen in OLAP

Thurnheer, Andreas. Temporale Auswertungsformen in OLAP. 2003, Doctoral Thesis, University of Basel, Faculty of Business and Economics.


Official URL: http://edoc.unibas.ch/diss/DissB_6469

Downloads: Statistics Overview


Online Analytical Processing (OLAP) enables end-users to define complex
multidimensional queries without the use of any programming skills. Most analytical
data models are based on the star schema, which differentiates between
dimensions and facts. While facts represent the main business measures, dimensions
contain the criteria to select, aggregate and navigate them. Since decisions
not only depend on current but also on historical data, facts are stored in
a time-dependent way. Although dimension data may also change over time,
many OLAP systems do not facilitate their time-dependent storage. As endusers
do not focus on dimension data in their analysis but on facts, it seems that
temporal aspects of dimension data can be disregarded at first view.
However, changes in dimension data may result in a situation, where a fact
value is assigned to different versions of a dimension object over time. In this
case, identical queries executed in different points of time possibly lead to unequal
results. Depending on the applied modeling technique and update strategy,
OLAP systems assign facts to the current, the original or to this version of
the dimension data which was valid at the time the facts were recorded. Since
OLAP systems can not update a data object with different methods at the same
time without having to maintain several data models simultaneously, it can
only provide one temporal perspective to dimension data.
This dissertation determines methods to enable the representation of different
temporal perspectives in an OLAP system. After the comparison of already
known modeling approaches, we introduce the Aggregated Fact Model. This
method extends a conventional OLAP system with a new class of temporal aggregates,
providing an alternative temporal perspective to dimension data. As
the definition, administration and analysis process is mostly driven by meta
data, neither administrators nor end-users will be faced with additional complexity.
Further Information can be found on the internet at the following address:
Advisors:Lusti, Markus
Committee Members:Rosenkranz, Friedrich
Faculties and Departments:06 Faculty of Business and Economics > Departement Wirtschaftswissenschaften
Item Type:Thesis
Thesis Subtype:Doctoral Thesis
Thesis no:6469
Thesis status:Complete
Number of Pages:188
Identification Number:
edoc DOI:
Last Modified:23 Feb 2018 11:40
Deposited On:13 Feb 2009 14:42

Repository Staff Only: item control page