The conventional rough set based feature selection techniques find the relevant features for the entire data set. However different sets of dimensions may be relevant for different clusters. This paper introduces a novel Projected Rough Fuzzy c-means clustering algorithm (PRFCM) which employs rough sets to model uncertainty in data, and fuzzy set theory to compute the weights of dimensions applicable to individual clusters. We discuss the convergence of the proposed algorithm and present the results of applying the proposed approach to several UCI data sets to demonstrate that it scores over its competitors in terms of several quality and validity measures.