Abstract—This paper presents a hybrid unsupervised clustering algorithm, referred to as the Rough Fuzzy C-Means (RFCM) algorithm and Particle Swarm Optimization (PSO). The PSO algorithm features high quality of searching in the near-optimum. At the same time, in RFCM, the concept of lower and upper approximation can deal with uncertainty, vagueness and indiscernibility in cluster relations while the membership function in a fuzzy set can handle overlapping partitions. To illustrate the competence of this method, a number of state-of-the-art hybrid methods (FPSO, Fuzzy-FPSO, RCM-PSO, K-means PSO) are compared through application on datasets obtained from the UC Irvine Machine Learning Repository. The reported results and extensive numerical analysis indicate an excellent performance on the proposed method.