05996384

a

05996384 Holst, Anders Ekman, Jan Incremental stream clustering for anomaly detection and classification. Kofod-Petersen, Anders (ed.) et al., Eleventh Scandinavian conference on artificial intelligence SCAI 2011. Papers based on the presentations at the conference, Trondheim, Norway, May 24--26, 2011. Amsterdam: IOS Press (ISBN 978-1-60750-753-6/hbk; 978-1-60750-754-3/ebook). Frontiers in Artificial Intelligence and Applications 227, 100-107 (2011). 2011 Amsterdam: IOS Press EN

doi:10.3233/978-1-60750-754-3-100

Summary: We have designed a framework for Bayesian Statistical Anomaly Detection, called ISC, or Incremental Stream Clustering. It learns the normal situation incrementally, and can on the fly detect anomalous cases. When this happens, a new cluster can be created, so similar cases can be detected in the future. In this way, the framework performs incremental clustering, while at the same time either classifying a new case as belonging to one of the known clusters or indicating that it is from a previously unseen situation.