<![CDATA[Gross Regional Domestic Product (GRDP) at constant prices (ADHK) is a key indicator for measuring real economic growth at the regional level. However, estimating GRDP at the regency/city level in Indonesia still faces challenges related to limited real-time data availability, publication delays, and reliance on conventional statistical methods that are often unable to capture complex and nonlinear relationships. This research aims to develop and compare several machine learning models in estimating ADHK GRDP for 27 regencies/cities in West Java Province using data from 2010–2024. The study employs a quantitative explanatory approach with panel data consisting of 405 observations obtained from the West Java Open Data portal. Feature engineering was conducted by incorporating historical growth rates, temporal variables, and regional encoding to capture temporal dynamics and spatial heterogeneity. Four predictive models were developed, namely linear regression, Random Forest, Gradient Boosting, and Support Vector Regression (SVR), and were evaluated using RMSE, MAE, MAPE, and R² metrics with cross-validation. The results indicate that ensemble-based models outperform traditional methods, with Gradient Boosting demonstrating the best performance by achieving the lowest error values and the highest explanatory power. Random Forest also shows strong predictive capability, while linear regression yields the lowest accuracy. These findings highlight the superiority of machine learning, particularly tree-based ensemble methods, in modeling complex regional economic data. The study contributes to the limited literature on regency/city-level GRDP estimation in Indonesia and suggests that machine learning can serve as a reliable tool for supporting data-driven policy formulation.]]>
