Mathematical techniques for summarizing large amounts of multidimensional data into groups. The two most popular techniques are:

hierarchical

k-means.

The hierarchical system calculates as many clusters as there are data points and displays their relative closeness by means of a dendogram. This system is preferred when there are few data points but the user wishes to see the dendogram to chose an appropriate number of clusters for analysis. Principal Component Analysis (PCA) is a form of hierarchical cluster analysis.
The k-means system requires the user to choose the number of cluster to be determined. The computation scatters the centers of the clusters among the data and then moves them until they are "gravitationally bound" to the larger groups of data and no longer move. The points determined in this way represent the central points of the clusters. This technique is very fast and appropriate for very large data sets. It is most commonly used in electrofacies calculations.
Cluster analysis is often used to provide electrofacies from wireline data where each curve is set to be a dimension.