Data Warehousing versus Event-Driven BI: Data Management and Knowledge Discovery in Fraud Analysis

M. Suntinger, J. Schiefer, H. Roth, H. Obweger:
"Data Warehousing versus Event-Driven BI: Data Management and Knowledge Discovery in Fraud Analysis";
Vortrag: International Conference on Software, Knowledge, Information Management and Applications, Kathmandu; 18.03.2008 - 21.03.2008; in:"International Conference on Software, Knowledge, Information Management and Applications", (2008), S. 44 - 49.

[ Publication Database ]

Abstract:


In the growing market of online betting and gambling,
fraud has reached a magnitude that cannot be overlooked.
For successful fraud detection and prevention, systems strongly
depend on a detailed characterization of recurring fraud patterns.
This paper compares two distinct technologies to extract this
knowledge: data warehouses and event-driven BI tools. Though
data warehousing coupled with OLAP analysis is in wide-spread
use, several limitations and shortcomings come to light. For
discovering fraud patterns, the abstraction of business events into
aggregated key figures makes detailed root-cause and cause-chain
analyses very difficult. Furthermore, complex data mappings
and comprehensive efforts for data integration are required for
preparing the data for analytical purposes in the data warehouse.
In comparison to data warehouses, event-based systems are easy
to integrate and provide the analyst with fine-grained information
for sound root-cause analyses. Business processes and behavioural
patterns of users can be fully reconstructed and visually analyzed
at the level of single events.