The Fundamental Clustering Problems Suite (FCPS) provides a set of clustering problems that an algorithm should be able to solve in order to be rightly called a clustering algorithm. FCPS can thus be regarded as an elementary benchmark for clustering algorithms.
FCPS contains data sets with known classification, which should be reproduced by the cluster algorithm. The data sets are deliberately kept simple and can be visualized in 2-3 dimensions. Each data set represents a specific problem, which is solved differently by known cluster algorithms. This allows the strengths and weaknesses of the algorithms to be identified. Standard clustering methods, such as single-linkage, Ward and k-means, are not able to solve all FCPS problems satisfactorily.
FCPS can and should be used freely in scientific papers, provided that the following reference is cited in publications and/or presentations:
Ultsch, A.: Clustering with SOM: U*C, In Proc. Workshop on
Self-Organizing Maps, Paris, France, (2005) , pp. 75-82
Click here for data (.zip)
Click here for a detailed description.