Model-Based Clustering of Microarray Data

Abstract

The K-means method (Hartigan and Wong 1979; Hartigan 1975; MacQueen 1967) is not a model-based clustering method. It is a data-partitioning method that divides the entire data into K disjoint groups. The idea is similar to the multivariate Gaussian mixture model analysis (a model-based method). Both deal with the problem of data partitioning and thus are in clear contrast to the hierarchical clustering methods described in Chap. 18. The K-means method is simple but very useful in microarray data analysis. As a result, it earns a section here in this chapter, even though it is not a model-based clustering method.