Drug Use Pattern Analysis Model Using the K-Means Algorithm as a Basis for Stocktaking Decision Making

Analysis of drug usage patterns using the K-Means algorithm aims to group drug usage based on its usage level and determine the most optimal number of clusters as a basis for more appropriate and efficient stock management recommendations. This study only focuses on drug usage data for the period of January 1, 2025, to June 30, 2025, with the variables used being the number of drugs used and their frequency of use. The approach used is data mining with the K-Means Clustering method, as well as cluster evaluation using the Elbow Method and Silhouette Coefficient. Using the Elbow method, the appropriate number of clusters is 3. This is indicated by the elbow point at k = 3, where the decrease in the WCSS value begins to decrease significantly. Evaluation of clustering quality using the Silhouette Score and Daevis-Bouldin Index (DBI) shows that the formed cluster structure has good quality. The Silhouette Score value reaches 0.61, and the DBI value is 0.53. This indicates that the data in each cluster is quite homogeneous, and the separation between clusters is quite optimal. The analysis results show that the most optimal number of clusters is three clusters, representing drug categories with high (fast-moving), medium (medium-moving), and low (slow-moving) usage levels. Each cluster has different but consistent usage characteristics. These findings provide a clear picture of the distribution pattern and drug needs at the Purwokerto Utara II Community Health Center,  and help identify the possibility of deadstock and stockouts. Thus, it can be concluded that the application of the K-Means algorithm is very effective in supporting drug stock management decision-making so that drug procurement planning can be carried out more accurately, efficiently, and sustainably.