<![CDATA[Indra Optik Blitar manages a large volume of patient medical record data that has not yet been utilized to support strategic decision-making. This study aims to cluster myopia refractive error patients based on their severity level, region, age, and gender characteristics using the K-Means algorithm, and to evaluate the quality of the resulting clusters using the Silhouette Coefficient. The research process follows the Knowledge Discovery in Database (KDD) framework, encompassing data selection, preprocessing, feature engineering, Min-Max Scaling normalization, and clustering execution. From 14,863 raw data entries, 6,821 valid records were obtained after three filtering layers: demographic, geographic, and clinical. Feature engineering was performed by converting four raw refraction variables into a single Worst_SE (Worst Spherical Equivalent) variable and transforming district address data into latitude and longitude coordinates. Feature weighting was automatically optimized using the Optuna framework with the Tree-structured Parzen Estimator (TPE) algorithm through 100 optimization trials. The optimal number of clusters was determined using a combination of the Elbow Method and Silhouette Coefficient, resulting in K = 5. The K-Means algorithm with K-Means++ initialization achieved convergence at iteration 20, producing five clusters with distinctive characteristics. Evaluation using the Silhouette Coefficient yielded an average score of 0.4144, which, given the high dimensionality and heterogeneity of the combined clinical, demographic, and geographic features, represents a moderately well-separated cluster structure suitable for practical application. The five clusters formed represent patient segments that differ clinically, demographically, and geographically, providing a data-driven foundation for Indra Optik Blitar to design more targeted marketing strategies, eye health socialization programs, and social assistance initiatives. ]]>
