Instruments for Cluster Analysis
Since our main job here is to decide cluster formation of a sample data set, we first have to measure the similarity. The most apparent answer for this question is to select distance metric d, where d can be defined in many ways. Here are some of the best known distance measures.
• Minkoski metric
• Euclidean metric
• Manhattan metric
• Mahalanobis Distance
Criterion functions for Clustering
As pointed out before, every
algorithm has criterion or cost function that is to be optimized. Every
clustering may have different outcomes depending on the objective function.
Determinant criterion and invariant criterion are the criteria that serve for
different clustering techniques.