Parameter selection in fuzzy joint points clustering algorithms

Applying fuzzy logic to clustering techniques leads to more robust and autonomous methods like the fuzzy joint points (FJP) which is a density based fuzzy clustering algorithm that requires no parameters to be set. However, a straightforward implementation of the method is rather slow. Recently, a faster but parameter dependent version of the algorithm was proposed and a theoretical bound on the parameter was given so that the algorithm produces the exact same results with the original FJP method. In this work, we investigate the tightness of the bound in practice and analyze the effect of the data distribution on the parameter selection problem of the fuzzy joint points clustering.