Bulletin
102, April 98
SVOR
EINLADUNG
Im Rahmen unserer Generalversammlung
organisieren wir ein öffentliches Seminar. |
ASRO
INVITATION
Dans le cadre de notre assemblée
générale, nous organisons un séminaire public. |
Programm / Programme
|
|
|
Freier Eintritt
Freitag, 8. Mai 1998
Bern, Hotel Alfa
Laupenstrasse 15, 3008 Bern |
Entrée libre
Vendredi, 8 mai 1998
Berne, Hotel Alfa
Laupenstrasse 15, 3008 Berne |
Heuristic Search Techniques in
Datamining
Dr. George D. Smith,
University of East Anglia, Norwich, UK
Datamining is the application
of algorithms to extract patterns from data and is one of the key parts
of knowledge discovery in databases (KDD).
In this context, datamining
can be viewed as an optimisation process in that one is trying to extract
the best pattern available from the data, usually in the form of rules.
Optimisation algorithms such as linear programming techniques have been
used for datamining. Here we concentrate on the application of heuristic
search techniques. A heuristic search technique is one that, although not
guaranteeing to find the optimal solution, will find good working solutions
in a reasonable time for problems in which exact optimisation techniques
may not be readily applicable, or for which such techiques will take too
long to find the optimal solution.
The presentation will firstly
describe the stages of datamining, then give a brief introduction to heuristic
techniques including genetic algorithms, simulated annealing and tabu search.
It will conclude with a case study on a real-world industrial application.
Data Mining: A Statistics
and Operational Research Perspective
Prof. Sally McClean,
Ulster University, Northern Ireland
As databases have become
larger there has been an accompanying explosion in the potential for Data
Mining, or Knowledge Discovery in Databases (KDD). This has involved developments
in machine learning, statistics and database theory, as components of large-scale
data analysis to extract new knowledge. The main developments which have
led to the emergence of Data Mining have been in the increased volume of
data now being collected and stored electronically, with an accompanying
maturing of Database Technology. Thus, whilst Data Mining research fuses
a number of deep academic research areas, it has also proved to be highly
relevant to many diverse application areas.
Many statistical methods
are being employed as part of Data Mining. In particular: statistical techniques
for data cleansing, exploratory data analysis, statistical methods for
data selection, statistical analysis methods, and probability models
such as Bayesian belief nets, evidence theory, and fuzzy systems.
Such techniques are being used to automate the discovery of patterns in
large and complex databases.
Business and Industry have
become major users of Data Mining methods. Such areas involve a wide range
of applications such as marketing, fraud and risk analysis, credit scoring
and customer profiling. Typically, we are concerned with decision making
which requires optimal utilisation of resources. By harnessing such large
volumes of data, using advanced methods of data access and management,
these Operational Research problems may utilise Data Mining methods to
provide useful and optimal solutions.
RETURN
to previous page.