<![CDATA[This study aims to develop a student potential clustering system as a strategy to address the limitations of academic identification processes that have traditionally been conducted manually, subjectively, and are prone to observational bias. The K-Means Clustering and K-Medoids algorithms were applied to a dataset consisting of 1,023 student records from SMP Negeri 2 Jekulo Kudus and SMP Negeri 3 Jekulo Kudus, using variables such as semester report card grades, core subjects including Mathematics, Science, and Indonesian Language, overall average scores, attitude assessments, and participation in extracurricular activities. The study employed a cluster number of (k = 3), representing High, Medium, and Low student potential categories for educational mapping purposes. The data preprocessing stage included missing value imputation using mean values and normalization of numerical features using RobustScaler to minimize the influence of outliers without removing student data. The evaluation results indicate that the K-Means algorithm achieved better clustering performance than K-Medoids based on evaluation metrics, with a Silhouette score of 0.529 and a Davies–Bouldin Index of 0.879, making it more suitable for the characteristics of the student dataset used. The system was subsequently implemented as an interactive web-based application developed in Python using the Flask framework and a MySQL database, enabling centralized data management, real-time access, and visualization of clustering results through a user-friendly interface. With this system, schools are expected to be able to map student potential more objectively, efficiently, and in a data-driven manner, thereby supporting learning strategy planning, intervention programs, and more targeted and inclusive educational decision-making.]]>
