<![CDATA[Diabetes risk monitoring often focuses on data recording and visualization, with limited analytical support for identifying interpretable risk patterns from health-related data. This study developed a web-based diabetes risk monitoring prototype that integrates K-Means clustering for analytical risk segmentation. Using the Pima Indians Diabetes Database with 768 records, this study applied invalid zero handling, median imputation, Min-Max scaling, K-Means clustering, internal validation, cluster profiling, and artifact preparation. Two preprocessing scenarios and candidate values of k were evaluated. The final model was selected under the Extended Scenario at k = 2 and produced two analytical risk segments: Higher Risk and Lower Risk. Cluster 0 contained 270 records (35.16%) and was labeled Higher Risk, while Cluster 1 contained 498 records (64.84%) and was labeled Lower Risk. The prototype was implemented using Flask, SQLite, HTML/CSS, Chart.js, and a saved model bundle. It supported data input, result display, prediction history, edit and delete functions, CSV export, and dashboard visualization. Black-box testing across 27 functional test cases showed that all tested features operated successfully. The main contribution is the integration of K-Means-based segmentation into an operational web prototype for structured monitoring, not clinical diagnosis.]]>
Copyrights © 2026
