Online one-class classification

                    Dr. David M.J. Tax
 
                   Fraunhofer FIRST.IDA
                 Kekulestr.7, D-12489 Berlin 
                          Germany                     

                         Abstract

In recent years the problem of one-class classification is recognized
as a new field of research. In constrast with normal classification
problems where one tries to distinguish between two (or more) classes
of objects, one-class classification tries to describe one class of
objects, and distinguish it from all other possible objects. This can
be used in novelty detection (for machine condition monitoring where
faults should be detected), outlier detection (for more confident
classification) or in badly balanced data (classification in medical
data with poorly sampled classes). In my talk I will present an online
version of a one-class classifier, the Support Vector Data Description
(SVDD), a classifier inspired by the support vector machines. By using
an online SVDD, dynamically changing data can be followed in time and
outlier objects can be detected. I will show some preliminary results
on the processing of EEG monitoring data.