Making Sense of Data: A Practical Guide to Exploratory Data Analysis and Data Mining

6.2 CLUSTERING

6.2.1 Overview

Clustering will group the data into sets of related observations or clusters. Observations within each group are more similar to other observations within the group than to observations within any other group. Clustering is an unsupervised method for grouping. To illustrate the process of clustering, a set of observations are shown on the scatterplot in Figure 6.5. These observations are plotted using two hypothetical dimensions and the similarity between the observations is proportional to the physical distance between the observations. There are two clear regions that could be considered as clusters: Cluster A and Cluster B. Clustering is a flexible approach to grouping. For example, based on the criteria for clustering the observations, observation X was not judged to be a member of Cluster A. However, if a different criterion was used, X may have been included in Cluster A. Clustering not only assists in identifying groups of related observations, it also locates observations that are not similar to others, that is outliers, since they fall into groups of their own.

Clustering has the following advantages:

• Flexible: There are many ways of adjusting how clustering is implemented, including options for determining the similarity between two observations and options for selecting the size of the clusters.
• Hierarchical and nonhierarchical approaches: Certain clustering techniques organize the data sets hierarchically, which may provide additional ...