Integrating GeoAI and Machine Learning–Based Geospatial Analysis for Data-Driven Territorial Decision-Making: A Quantitative and Spatial Modeling Approach

Abstract

The integration of geospatial analysis and Artificial Intelligence (AI) offers new opportunities to improve territorial decision-making through predictive, data-driven approaches. This study presents a quantitative and explanatory framework that combines Geographic Information Systems (GIS) with machine learning models to analyze spatial patterns and assess territorial risk. The analysis is based on real geospatial data obtained from secondary sources, including administrative spatial units, satellite-derived indicators, and publicly available socio-territorial datasets.

The methodology follows a transparent and reproducible workflow that includes spatial data preprocessing, exploratory spatial data analysis (ESDA), and the implementation of GeoAI models, specifically random forests and artificial neural networks. Model configuration, validation strategies, and performance metrics are explicitly defined and compared with conventional GIS-based regression approaches. Spatial autocorrelation is assessed using Moran’s I and LISA statistics, and the results are visualized through spatial maps to support territorial interpretation.

The findings indicate that AI-enhanced geospatial models significantly outperform traditional GIS methods in terms of predictive accuracy, spatial precision, and explanatory power. The improved identification of high-risk areas demonstrates the practical value of GeoAI for territorial planning, resource allocation, and policy design. This study contributes to the growing field of GeoAI by providing a methodologically explicit and policy-relevant framework that supports transparent, reproducible, and evidence-based territorial decision-making.