Hi Joel Luis Carbonera : There are a variety of indices used to evaluate the strength of a clustering technique. For example, the Silhouette index is a widely used metric that measures the average separation of nodes belonging to two different clusters as compared to the average of separation of nodes within the same cluster. A variety of different indices for evaluation have been discussed in the attached paper.

Hi Joel Luis Carbonera,

I want to clarify that, the cluster analysis is too much depended on the data characteristic, Even in given clustering method several factors such as: number of clusters, similarity function selection, desired accuracy can change the performance of the clustering.

If you have pre-knowledge about the clustering results (i.e. using simulated data for testing the performance of clustering) you may try different clusterings to find more accurate one. This can help you to know which method works better than others (i.e. other reported data is also good for testing).

Some packages are available in R which can help you to detect optimal number of clusters such as "NbClust" package see : https://www.rdocumentation.org/packages/NbClust/versions/3.0/topics/NbClust

I hope it helps,

Reza

Cheers,

Dear Mahini

It depends on your data. I should suggest you to examine several method on a part of the data. Then, you can decide that which of them is proper for your project.

Further, some preprocessing approaches may be useful before applying the method.