Yin Zhao, George Karypis, Criterion Funtions for Document Clustering: Experiments and Analysis, Technical Report #01-40, 2002.

WORK IN PROGRESS...

DESCRIPTION: In this paper, clustering is viewed as an optimization problem which seeks to maximize or minimize a particular criterion funtion; some algorithms in this class are K-means and Autoclass. Seven criterion functions are investigated across multiple datasets and two partional methodologies: direct k-way clustering and via repeated bisections. Documents are represented using the Vector Space model with tf-idf weighting. The seven criterion functions are divided in Internal, External and Hybrid. The Internal functions maximize over the documents that are part of a ...

COMMENTS: It only deals with off-line clustering.

-- AlvaroBolivar - 11 Jul 2003
to top