In many fields of empirical research, data sets
to do not only become larger in size, but also in complexity. Many standard
statistical techniques used for analyzing this type of data are not resistant in
presence of outliers and become invalid. While robust methods are well
established for dealing with simple models, as the regression and location-scale
model, there is still work to do for more complicated, multivariate and
non-linear models. Since atypical observations are frequently present when
analyzing complex data sets, new robust methods need to be introduced.
In this track session we would like to focus on methods that are considered as
data-mining techniques, including supervised and unsupervised learning. In the
field of supervised learning, with discriminant analysis and logistic regression
as oldest representatives, it is common to have a large number of possible
predictor variables. Robust variable selection methods need to be developed here.
Also, reliable and robust prediction methods need to be worked out. Another
issue is to study the robustness properties of well known methods as
classification trees methods. Those are often claimed to be robust, but have not
much been studied by researchers from the robustness community. Practical
implementation and computational feasibility are of major importance in robust
data mining.