Comparing K-Means and K-medoids algorithms for clustering hamlet regions by tax liabilities in tax determination documents

The application of data mining information technology in Village Offices, especially in village office administration services, is very important to ensure the efficiency and accuracy of information services. This research aims to compare the effectiveness of the K-Means and K-Medoids algorithms in clustering hamlet areas based on the tax owed in the tax assessment documents in Pandanarum village. Using quantitative descriptive methods, the two algorithms are applied to group hamlets based on tax payable data as the main variable. The clustering process is analyzed using evaluations such as Sum of Squared Errors (SSE) and Silhouette Score to determine the effectiveness of each algorithm. The research results show that the K-Medoids algorithm has lower performance compared to K-Means, especially in terms of cluster stability and a high Silhouette Score value with a value of 0.454615 and SSE 480.9462. Apart from that, the K-Medoids algorithm is more robust against outliers in the tax payable data, and produces a lower Silhoutte Score value with a value of 0.382616 and an SSE of 567.6125 which indicates weaker clustering. Thus, this research concludes that the K-Means algorithm is superior in clustering hamlet areas based on taxes owed compared to the K-Medoids algorithm.