<![CDATA[This study evaluates the predictive performance of remote sensing variables, socioeconomic indicators, and their integration for estimating regional economic growth across 33 districts and municipalities in North Sumatra, Indonesia. A quantitative cross-sectional design was employed using three predictor scenarios: remote sensing variables (Nighttime Light, NDVI, Built-up Area, and Land Surface Temperature), socioeconomic indicators (Human Development Index, Open Unemployment Rate, Fiscal Capacity Index, and Disaster Risk Index), and an integrated dataset. Four regression algorithms (Linear Regression, Support Vector Regression, Random Forest, and K-Nearest Neighbors) were optimized using RandomizedSearchCV and validated through Leave-One-Out Cross Validation. Model performance was evaluated using RMSE, MAE, and R², while permutation importance assessed predictor contributions. Support Vector Regression achieved the best predictive performance across all predictor scenarios. The socioeconomic dataset yielded the highest prediction accuracy (RMSE = 0.625, MAE = 0.480, R² = 0.198), outperforming the remote sensing-only dataset (RMSE = 0.685, MAE = 0.471, R² = 0.038) and the integrated dataset (RMSE = 0.647, MAE = 0.492, R² = 0.140). Although the R² values were relatively low, they reflect the complexity of regional economic growth and the influence of factors beyond those included in this study. Permutation importance identified the Human Development Index and Disaster Risk Index as the most influential predictors. These findings indicate that socioeconomic indicators are stronger predictors of regional economic growth, while remote sensing variables provide complementary spatial information. Although integrating remote sensing variables did not improve predictive accuracy, it offers valuable environmental context for more comprehensive data-driven regional development planning.]]>
Copyrights © 2026
