for the give data 100*2 how many clusters will get. is there any relation with any parameter r any formula?

based on the obtained data how to predict clusters?

Answer 1

[Ameer Hamza]

The problem of choosing the correct numbers of clusters is quite subjective and depends on your application requirement. You can refer to these links to get some idea about how to determine an optimal number of clusters from the dataset:

https://en.wikipedia.org/wiki/Determining_the_number_of_clusters_in_a_data_set

http://www.sthda.com/english/articles/29-cluster-validation-essentials/96-determining-the-optimal-number-of-clusters-3-must-know-methods/

https://stats.stackexchange.com/questions/23472/how-to-decide-on-the-correct-number-of-clusters

MATLAB clustering algorithm e.g. kmeans() allows to cluster data once you know the number of the cluster in which you want to divide the data.

[John D'Errico]

Exactly. It can be difficult to know in advance how many clusters you need. If there were an easy way, using some magic formula, the code would use it, or at least offer such an option.

[Walter Roberson]

In the past I had occasion to test several algorithms to determine the number of clusters. None of the ones I tested produced good results. They depended on arbitrary tuning parameters whose value could not be justified, or they were sensitive to small accidents of point placement, or they would say that 27 clusters should be used instead of 4 because they had weak penalties for generating more clusters (best fit for clustering us always one cluster per unique point unless there are penalties against cluster creation.)

The ones with tunable parameters... it wasn't that they didn't work (in some sense), but rather that we were trying to produce unsupervised learning for situations where we would have no reason to expect that clusters would have a particular shape. So the only effective way to use the algorithm was to run the tunable parameters themselves through a search algorithm...