Practical Predictive Analytics

What is a cluster?

A cluster can be defined as a group of objects with similar characteristics. Clusters can vary in size. There is no requirement that a cluster must have a certain number of members. It can even have a single member. This might be useful in fraud applications, in which you want to identify anomalies in the data. In this case, very small clusters can act as outliers.

Clusters are often constructed to have an equal number of members in each group, as in partitioning clustering. In this case, all of the members in each cluster are homogeneous within each cluster group.

Furthermore, clusters are constructed so they have distinct differences from the remaining clusters. Differences are usually calculated based upon measurements ...