Clustering isi a method fer partitioning a set of objects into homogeneous groups (clusters) based on a specified set of variables. goals of this method is objects whitin a cluster are similiar and dissmilar with the objects in other clusters. K-Harmonic means (KHM) is a clustering algorithm that can solve problems on the cluster center intialization of K-Means algorithm, but KHM still ca not overcome local optima problem. particle swarm optimization (PSO) is a stochastic algorithm that can used tio find optimal solution to a numerical problem, but PSO has a problem at the convergence speed. to overcome these problems, there is particle swarm optimization K-Harmonic means (PSOKHM) algorithm which is a combination of KHM and PSO algorithm. in this final project, PSOKHM algorithm used to perform data clustering, and KHM and PSO algorithm as a comparison for evalution of the cluster-based objective function value, F-Measure, and the runnning time. trials conducted with 3 scenarios of 5 differnt data sets. from the result of the test obtained that, when viewed from the objective function and F-Measure value, PSOKHM able to give better. meanwhile, if viewed from the running time, PSOKHM surpasses PSO but is not better than KHM.