SODA

Incremental stream clustering and anomaly detection

Ekman, Jan and Holst, Anders (2008) Incremental stream clustering and anomaly detection. [SICS Report]

[img]
Preview
PDF
799Kb
[img]
Preview
Postscript
6Mb

Abstract

This report concerns the "ISC-tool", a tool for classification of patterns and detection of anomalous patterns, where a pattern is a set of values. The tool has a graphical user interface "the anomalo-meter" that shows the degree of anomaly of a pattern and how it is classified. The report describes the user interaction with the tool and the underlying statistical methods used, which basically are Bayesian inference for finding expected or "predictive" distributions for clusters of patterns and using these distributions for classifying and assessing a degree of anomaly to a new pattern. The report also briefly discusses what in general are appropriate methods for clustering and anomaly detection. The project has been supported by SSF via the Butler2 programme.

Item Type:SICS Report
Uncontrolled Keywords:Incremental Clustering, Anomaly detection, Bayesian Statistics, Classification
ID Code:2901
Deposited By:Vicki Carleson
Deposited On:24 Apr 2008
Last Modified:18 Nov 2009 16:14

Repository Staff Only: item control page