Customer segmentation is an important strategy in a company, it affects good customer relationships which will result in increased profits. Grouping customers in data mining can use several algorithms, but K-Medoids is the right choice because it can reduce noise and outlier sensitivity. However, the selection of cluster centers is still random and has an effect on the results of clustering, so it is necessary to improve the k-medoids algorithm so that the resulting cluster value can be optimal. Particle Swarm Optimization is an optimization algorithm that is often used and has been proven to improve the results of a clustering. In this case, optimization using Particle Swarm Optimization (PSO) in the selection of the initial cluster center needs to be applied to the k-medoids algorithm so that the results of the cluster can be optimal. The results of the study showed the Davies-Bouldin Index (DBI) value for K-Medoids K 2 = 0.379, K 3 = 0.283, and K 4 = 0.593, while the DBI value PSO + K-Medoids K 2 = 0.088, K 3 = 0.226, and K4 = 0.363. The DBI value shows that PSO optimization on K-Medoids to determine the initial centroid is proven to improve the results of clustering than standard K-Medoids.