Integration of Geo-Spatial Data and Machine Learning for Socio-Economic Forecasting

Abstract

This study proposes an integrated Geo-Spatial Artificial Intelligence (GeoAI) framework that combines geospatial data and machine learning techniques to forecast socio-economic indicators at the regional level. The primary objective is to generate accurate and spatially informed predictions of socio-economic conditions to support evidence-based regional planning and policy development. The study adopts an applied research design by integrating geospatial datasets derived from satellite imagery, infrastructure databases, and socio-economic statistics. Key geospatial variables include vegetation density (NDVI), nighttime light intensity, infrastructure accessibility derived from road networks, population density, and built-up area indicators. These spatial attributes are combined with demographic and regional data through spatial data processing techniques such as spatial joins and grid-based spatial aggregation. To capture complex relationships between geospatial variables and socio-economic conditions, supervised machine learning algorithms were implemented, including Random Forest and Gradient Boosting models. These algorithms are widely used in GeoAI applications due to their ability to model nonlinear relationships and handle heterogeneous spatial datasets. An ensemble prediction approach was applied to improve model robustness and prediction accuracy. Model performance was evaluated using regression-based metrics including Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). In addition, spatial autocorrelation analysis using Moran’s I was conducted to assess spatial dependency in the distribution of socio-economic indicators. The experimental results demonstrate that the proposed GeoAI framework successfully captures spatial patterns of socio-economic indicators across districts. Feature importance analysis indicates that nighttime light intensity, population density, and infrastructure accessibility are among the most influential predictors. Spatial visualization of prediction results highlights clear regional disparities, where urban districts tend to exhibit higher predicted socio-economic scores compared to rural areas with lower infrastructure accessibility. These findings confirm that the integration of geospatial analytics and machine learning significantly enhances the ability to model and forecast socio-economic conditions. The proposed framework provides a scalable and data-driven approach for regional socio-economic analysis and offers valuable insights for infrastructure planning, resource allocation, and sustainable regional development. Future research may further improve the robustness of the GeoAI framework by incorporating spatial cross-validation techniques and additional geospatial variables to better capture spatial dependencies across regions.