<![CDATA[Dengue fever remains a pressing public health challenge in major Indonesian cities, including Semarang. The complex interplay of heterogeneous demographic structures and built-environment characteristics generates spatially uneven transmission risks, while conventional risk-mapping approaches often fail to capture the probabilistic nature of these risks at fine-scale administrative levels, limiting their utility for targeted interventions. This study aims to develop a robust, replicable framework for dengue risk stratification that more accurately identifies localized high-risk areas and supports evidence-based public health decision-making. The research introduces a probabilistic clustering approach using Gaussian Mixture Models (GMM) to move beyond rigid partitioning methods, while simultaneously integrating multi-year incidence data (2021–2024) with eighteen multidimensional urban indicators across 177 sub-districts (kelurahan). This combined contribution advances methodological rigor by accommodating overlapping data distributions and probabilistic cluster memberships, and provides a nuanced, evidence-driven tool for stratifying dengue risk and guiding hyper-local interventions. Several GMM configurations were evaluated using the Bayesian Information Criterion (BIC) to determine the optimal number of clusters. The BIC value declined markedly when the number of clusters increased from two to three, indicating a substantial improvement in model fit. Further increases yielded only marginal gains, and the lowest BIC was achieved at three clusters, representing the most parsimonious and effective solution. Internal validation confirmed that the cluster structure robustly captured epidemiological variance despite the inherent heterogeneity of urban spatial data. Cluster 2 emerged as a critical high-risk epicenter, geographically limited yet characterized by consistently elevated incidence, pronounced temporal variability, and extreme values. The proposed GMM-based framework demonstrates that dengue risk in Semarang is concentrated within localized foci of heightened vulnerability rather than uniformly distributed. Ultimately, the methodology is replicable in other complex tropical urban environments, thereby strengthening both academic rigor and practical public health decision-making]]>
