The K-Medoids method is a non-hierarchical cluster analysis method where information on the exact number of clusters is required. The data used in this study uses simulation data from reference data on the percentage of households according to drinking water sources. The simulation data used uses a multivariate normal distribution, so that the simulation data allows for negative data. In this study, two options were carried out on negative data results, namely being zero and absolute. The method in determining the optimal number of clusters used the Sillhouette Coefficient method, the Elbow method and the Gap Statistics method. The average Dunn Index value from the data on the zeroed option produces the largest Dunn Index value in determining the optimal number of clusters using the Gap Statistic method, which is 0,125734, while in the second option data, the Dunn Index average is greatest in determining the number of clusters optimally using the Sillhouette Coefficient method, which is 0,113315.