HPL-2012-136On the Helmholtz Principle for Data Mining
Dadachev, Boris; Balinsky, Alexander; Balinsky, Helen; Simske, Steven
Keyword(s): Unusual Behaviour Detection; Keywords Extraction; Summarization; Helmholtz Principle; Small-World Networks
Abstract: Unusual behaviour detection and information extraction in streams of short documents and files (emails, news, tweets, log files, messages, etc.) are important problems in security applications. In , , a new approach to rapid change detection and automatic summarization of large documents was introduced. This approach is based on a theory of social networks and ideas from image processing and especially on the Helmholtz Principle from the Gestalt Theory of human perception. In this article we modify, optimize and verify the approach from ,  to unusual behaviour detection and information extraction from small documents.
External Posting Date: June 28, 2012 [Abstract Only]. Approved for External Publication - External Copyright Consideration
Internal Posting Date: June 28, 2012 [Fulltext]