This study incorporates a spatial clustering technique into the formation of a spatial weight matrix as an alternative to the traditional exogenous matrix, aiming to better capture spatial dependencies. The approach is applied to analyze the spatial autocorrelation of economic growth in East Java’s regencies and municipalities using 2019–2021 data. Spatial clusters are identified based on GDP growth (GGDP), Human Development Index (HDI), population density (Dens), and geographical coordinates. These clusters are used to define a customized spatial weight matrix, where regions within the same cluster are designated as neighbors. Moran’s I, calculated using the customized spatial weight matrix, detects significant spatial autocorrelation in GDP growth for all three years, with consistently lower p-values compared to the traditional contiguity-based matrix. For example, in 2020, Moran’s I using the customized matrix yielded a p-value of 0.099 (significant at the 10% level), while the traditional matrix produced a non-significant p-value of 0.7965. These results demonstrate that spatial clustering extends the scope of spatial interaction beyond adjacent regions to include those with similar characteristics. The findings highlight the effectiveness of this method in providing a more nuanced and robust framework for analyzing spatial dependencies in economic growth.