Correlation-Based Cluster Analysis Using Mixture of Constrained PCAs

Taku Yoshioka, Natsuko Kawase, Shin IshiiProceedings Genome Informatics Workshop/Genome Informatics 2002 · DOI

Abstract

In a cluster analysis, similarity measure must be appropriately specified. When dealing with gene expression time-series data, it is important for similarity measure to put more emphasis on the temporal behaviors than on the scale (magnitude) of the data. The correlation coefficient is one of the most popular similarity measure, which is insensitive to the scale. It is often used in hierarchical clustering. However, there is no theoretical way to determine the number of clusters from results from hierarchical clustering. In this report, we propose a probabilistic model for correlation-based cluster analyses. The probabilistic formulation makes the determination of the cluster number a model selection problem, and the problem is statistically solved within the Bayesian framework.
BibTeX
@inproceedings{yoshioka2002correlation,
  author = {Taku Yoshioka and Natsuko Kawase and Shin Ishii},
  title = {{Correlation-Based Cluster Analysis Using Mixture of Constrained PCAs}},
  booktitle = {Proceedings Genome Informatics Workshop/Genome Informatics},
  year = {2002},
  doi = {10.11234/gi1990.13.256},
}

Loading related content...

Figure