ECML PKDD 2006 Workshop on
Practical Data Mining: Applications, Experiences and Challenges

Abstracts

Stefan Wrobel, Fraunhofer IAIS & University of Bonn

Every customer has an address, every store has a location, and traffic networks are a decisive factor in accessibility and logistics. Even in classical business data analysis, a large majority of data have a spatial component, and optimal business decisions must take geographical context into account. In the talk, we will present several examples of real world customer projects ranging from location selection and geo-marketing to outdoor media. We will then move on to the new challenges and opportunities brought about by the widespread availability of localisation technology that allows tracking of people and objects in time and space.

Professor Dr. Stefan Wrobel is a professor of computer science at university of Bonn and one of the three directors of the Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS (created in July 2006 as a merger of Fraunhofer Institutes AIS and IMK).

Directions of Analytics, Data and Text Mining - A software vendor's view

Ulrich Reincke, Executive Expert Analytics, SAS Institute Germany

After the years of hype at the turn of the millennium, immediately followed by the crush of the dot.com-bubble, data mining has become a mature market. New business applications are continuously developed even for remote industries and total new data sources are becoming increasingly available to be explored with new Data Mining methods. The common type of data sources moved initially from numerical over time-stamped to categorical and text, while the latest challenges are geographic, biological and chemical information, that are both of text and numerical type coupled with very complex geometric structures.

If you take a closer look at the concrete modelling options of both freeware and commercial data mining tools, there is pretty little difference between them. They all claim to provide their users with the latest analysis models that are consensus within the discussions of the research community. However, what makes a big difference, is the ability to map the data mining process into a continuous IT-flow, that controls the full information from the raw data, cleaning aggregation and transformation, analytic modelling, operative scoring, and last but not least final deployment. This IT process needs to be set up as to secure that the original business question is solved and the resulting policy actions are applied appropriately in the real world. This ability constitutes a critical success factor in any data mining project of larger scale. Among other environmental parameters of a mining project it depends mainly on clean and efficient metadata administration and the ability to cover and administer the whole project information flow with one software platform: data access, data integration, data mining, scoring and business intelligence. SAS is putting considerable effort to pursue the development of its data mining solutions in this direction. Examples of real life projects will be given.

Ulrich Reincke
Executive Expert Analytics,

Competence Center Enterprise Intelligence
SAS Institute, Heidelberg
In der Neckarhelle 162
69118 Heidelberg
Germany
Tel +49 (0)6221 415-2150
Ulrich.Reincke@ger.sas.com