Determining the number of clusters in the statistical data

Boldak A.A., Suharev D.L.

In the work three heuristic algorithms for automatic determination of the clusters amount of data samples were developed and experimentally studied. Usage of them can improve well-known clustering algorithms. Feature of the proposed algorithms is that there is no need to execute multiple calculations of clustering task for different clusters amount, with subsequent analysis of cluster structure quality.

Full text (pdf)