are there way to know the clusters are meaning or meaningless?

hi, Are there way to know if the algorithm of clustering give meaning or meaningless clusters?

thanks in advance

Answer 1

[Walter Roberson]

If you are doing k-means or other clustering that starts from random initial conditions, then you repeat the clustering a number of times with different conditions; the clusters that appear in more of the results are more probable.

If you are trying to determine the optimal number of clusters, then you have a Problem unless you can write down exactly what it means for a clustering to be optimal in your situation; there is no generally applicable definition of "optimal" that can be used.

[huda nawaf]

thanks , I'm not talk about optimal, but need to be sure if the clusters that result from my algorithms are meaning.

I used several types of hierarchical clustering such as ward, single, complete,... that built in matlab , in additin I built agglomerative clustering and newman algorithm . regarding initial condition , there is no initial conditions in these types. what I did is trying to see shared clusters among these methods, and I'm not sure if it is good way to cluster my data or not. I think there is some way more efficiecnt to do that.

regarding no. of clusters, I read about elbow as good way to determin the number of clusters, but it is not used with hierarchical method

someone commented about that in this link http://www.mathworks.com/matlabcentral/newsreader/view_thread/280153

he say that we can determine the number of clusters by looking at the similarity measures in hierarchical clustering. unfortunately, there is no way to cantact him

[Walter Roberson]

It is not possible for a computer program to determine whether a given output has meaning. An output is an output is an output is an output. Computers follow instructions, not assign meaning.