Biomedical Engineering Reference
In-Depth Information
Cluster analysis may reveal associations and structure in data that, though not previously evident,
are sensible and useful once found. The results of cluster analysis may contribute to the definition of
a formal classification scheme, such as a taxonomy for related bacteria. It may suggest statistical
models with which to describe populations, or indicate rules for assigning new cases to classes for
identification and diagnostic purposes.
Cluster analysis includes metric-, model-, and partition-based methods (see
Figure 7-11
). In metric-
based clustering, the data are partitioned so that they are closer to the centroid or center of mass
than they are to other data in the cluster. In model-based clustering, a hypothetical model for each
cluster is defined and the data that best fit the model are considered part of that cluster. A problem
with model-based approaches is overfitting—by chance, a model may fit data that is irrelevant to it.
Partition-based methods, which are general cases of metric- and model-based methods, use an ad
hoc method of dividing the data space.
Figure 7-11. Cluster Analysis Methods.
