<![CDATA[Clustering, an unsupervised machine learning technique, categorizes objects into groups based on shared characteristics. When applied to spatial data, the assumption of independence is often violated due to similarities among adjacent regions—a phenomenon known as spatial autocorrelation. To address this, spatial clustering incorporates both non-spatial attributes (e.g., socio-economic indicators) and spatial attributes (e.g., geographic location), with spatial attributes weighted based on their influence in defining clusters. In regional economic development, creating clusters that are both spatially coherent and socio-economically homogeneous is critical for effective policy design. Strong interactions among neighboring regions can promote more integrated and balanced growth. This study proposes a spatial clustering framework that optimizes spatial attribute weighting according to the degree of spatial autocorrelation. A simulation study using 2023 data from East Java’s 38 regencies/municipalities determines optimal weights under varying spatial dependence levels. The results show that optimal spatial weights increase with the number of clusters and vary according to the strength of spatial autocorrelation. Applied to East Java, the method produced clusters with higher socio-economic homogeneity than official zones, though with reduced spatial contiguity. These findings highlight the importance of adaptive, autocorrelation-aware clustering to improve regional planning and support more evidence-based development strategies.]]>
Copyrights © 2026
