SUMMARY
Abstract Merging several government agencies into one resulted in employee collaboration. The freedom for employees to choose a new work unit that matches their qualifications and domicile results in an unequal distribution of employees. Acceptance of CASN (Prospective State Civil Apparatus) is one of the efforts to cover the shortage of staff so that the number is proportional to the needs of the work unit. The purpose of this research is to cluster employee density in government agencies with the K-Means algorithm using the attributes of the number of employees according to qualifications and employees approaching retirement age, so that work units can be clustered from overcrowded to understaffed. The K-Means algorithm is a data mining method that performs an unsupervised modeling process and is a method for grouping data with a partition system. In this study, employee density clustering was formed into 4 clusters (k = 4), namely C1 (Dense), C2 (Moderate), C3 (Less), and C4 (Very Less). Clusterization testing using the RapidMiner software with the Davies Bouldin Index (DBI) method resulted in an accuracy value of 0.299. The results of the clustering of employee density from 84 work units in a government agency found that 19 work units belonging to cluster C4 were severely understaffed. The results of this clustering can be used to support decision making in predicting CASN needs in government agencies. Keywords — clustering, K-Means, Davies Bouldin Index, Data Mining