A Scalable Clustering–GIS Decision Support System for Municipal Early-Warning Governance: Design Science Research in a Health Smart City Context

Abstract

Local governments rely on integrated information systems to shift public-sector decision-making from reactive to proactive, yet the design of scalable, clustering-based spatial decision support systems (SDSS) for city-level early warning remains under-explored within Health Smart City governance. Following a Design Science Research approach, this study designs, builds, and demonstrates a Decision Support System (DSS) that integrates the K-Medoids CLARA clustering algorithm with a Web-based Geographic Information System (GIS) to generate spatially differentiated early-warning alerts, using Non-Communicable Disease (NCD) risk surveillance in Bogor City as the application domain. The artifact was instantiated on 26,798 electronic medical records from 25 Puskesmas via a client–server architecture comprising data anonymization, clustering, spatial aggregation, and an interactive alerting dashboard, and evaluated against clustering-quality and comparative-performance criteria. CLARA (k=3) achieved a Silhouette Coefficient of 0.2171 and Davies-Bouldin Index of 1.4977, modestly higher than K-Means on both metrics, while remaining scalable for large administrative datasets where standard K-Medoids (PAM) becomes impractical. The resulting four-level dashboard classified 4 of 25 Puskesmas as Level 3 (Critical), 11 as Level 2, 6 as Level 1, and 4 as Level 0, with GIS visualization communicating spatial risk to non-technical policymakers. The findings demonstrate a design-and-demonstration proof-of-concept showing how a clustering-GIS DSS architecture can be built and evaluated for technical feasibility, offering a transferable design pattern for public-sector information systems that integrate large-scale operational data with spatial early-warning capability; organizational adoption and field deployment remain future work.