<![CDATA[Sub-district socio-economic data on family welfare from BKKBN is high-dimensional and multivariate. The data pattern is highly complex with a positive asymmetric (right-skewed) distribution. This study aims to implement the Self-Organizing Maps (SOM) algorithm to cluster these welfare characteristics, mathematically evaluate model performance to eliminate bias, and visualize the tabular results into a spatial digital map. A 2×2 SOM model optimized using Optuna was used to group family welfare characteristics in West Java for the 2022–2025 period. Optimization results identified the best hyperparameter combinations (learning rate 0.7074–0.9574, sigma 0.1167–0.2668, and 7,000–20,000 iterations). The evaluation demonstrated reliable performance with Quantization Error ranging from 2.4659–3.4776, Topographic Error between 0.0000–0.0647, and Silhouette Score consistently above 0.50. GeoJSON integration successfully mapped 4 welfare hierarchies, confirming the dominance of prosperous regions in the eastern, southern, and western areas.]]>
Copyrights © 2026
