<![CDATA[Urban Heat Island (UHI) is a phenomenon of increased urban temperatures that affects thermal comfort, energy demand, and environmental quality. In Indonesia, UHI studies are still dominated by satellite-derived land-surface temperature, whereas continuous in-situ meteorological observations are less frequently used for multivariate data-driven grouping. This study addresses that gap by implementing K-Means clustering to identify station-based UHI-related thermal regimes in Banten Province using observations from 17 Automatic Weather Stations (AWS) during 2024. The source data were recorded at 10-minute intervals and processed through range and step quality-control checks, removal of invalid observations, Multiple Imputation by Chained Equations (MICE) for missing-data treatment, station-level aggregation, and Min-Max normalization. The final clustering features were annual mean air temperature, relative humidity, and solar radiation; rainfall was excluded from clustering because its correlations with the principal thermal variables were comparatively weak. Candidate cluster counts from K = 2 to K = 5 were evaluated using the Elbow method, Silhouette Coefficient Index (SCI), and Davies-Bouldin Index (DBI). Although the Elbow curve shows a structural transition around K = 3, internal validation identifies K = 2 as the most defensible final solution, with SCI = 0.79 and DBI = 0.33, compared with SCI = 0.65 and DBI = 0.58 for K = 3. The selected solution separates a higher-thermal regime concentrated in the northern and eastern urban-industrial corridor from a lower-thermal regime concentrated mainly in southern and western Banten. Because the AWS network is spatially sparse, the result should be interpreted as station-based thermal-regime clustering that complements, rather than replaces, spatially continuous remote-sensing UHI analysis. The resulting framework can support targeted heat monitoring and provide a data-driven basis for subsequent urban-climate assessment.]]>
