The literature encompasses various approaches to address sparsity in the context of cluster analysis, often by adding regularization terms to weight the role of variables in clustering processes. In the work presented here, instead, an L \(_0\) -regularization term is employed on fuzzy membership degrees to enhance the standard Fuzzy K-Means algorithm. The new algorithm helps to assign units very close to the corresponding prototype with a membership degree equal to 1 without necessarily compromising the potential ambiguity in the membership of some units.