Machine Learning with Spatial Context: Comparing Bayesian Geoadditive Regression Tree, Random Forest, and Gradient Boosting Machines for Gross Domestic Regional Product in Indonesia

This study aims to evaluate and compare the performance of five machine learning algorithms, namely Bayesian Additive Regression Trees (BART), Bayesian Geoadditive Regression Trees (BART Spatial), Decision Tree, Random Forest, and Gradient Boosting Model (GBM) in predicting Regional Domestic Product (RDP) in Indonesia based on spatial data. The main objective of this study is to assess the effectiveness of each model in predicting regional economic growth, focusing on the use of spatial data that describes each province's geographical and economic characteristics. The dataset used includes data from all provinces in Indonesia in 2024, obtained from the Indonesian Central Statistics Agency. The five machine learning models were trained using 80 % of the data and validated using the remaining 20%. Model performance was evaluated using Root Mean Square Error (RMSE) and the coefficient of determination (R-squared). The study results indicate that the boosting-based approach (GBM) outperforms other methods in predicting RDP, demonstrating significant potential in forecasting regional economic growth. These findings can assist policymakers in designing more targeted interventions to reduce economic disparities between regions. This study also contributes to developing more accurate economic prediction models by considering spatial factors, which have not been widely applied in Indonesia's GRDP context.