Estimación del valor de apartamentos en la localidad de Usaquén, Bogotá, mediante un enfoque de inteligencia artificial geoespacial, GeoAI

Abstract

The increasing availability of geospatial, real estate, and digital data sources has transformed urban market analysis by enabling Geospatial Artificial Intelligence (GeoAI) to integrate spatial analytics, machine learning, and unstructured data processing into real estate valuation. Traditional mass valuation methods are predominantly based on physical and regulatory attributes, scarcely incorporating spatial heterogeneity and territorial variation in the influence of variables on property prices. This problem is particularly evident in the Usaquén district of Bogotá, D.C., due to the coexistence of multiple socio-spatial dynamics. This study developed a four-stage methodological framework that integrates Geographic Information Systems, spatial econometrics, Natural Language Processing, and machine learning to estimate the market value of apartments. This framework considers the integration of heterogeneous data, the engineering of accessibility and urban environment features, the construction of a Spatial Weighted Sentiment Index (SWSI) extracted from georeferenced textual information, and the comparative evaluation of Ordinary Least Squares, Spatial Error, Geographically Weighted Regression (GWR), and Random Forest (RF) models. The results confirmed a significant spatial autocorrelation in the distribution of prices and identified that, while the GWR offered interpretability on the local variation of the coefficients, Random Forest demonstrated a greater aptitude for capturing non-linear relationships and the incorporation of the SWSI provided a qualitative dimension on the perception of the environment, demonstrating that the integration of predictive spatial analytics and citizen perception data generates spatially coherent value surfaces with high market representativeness, constituting a transferable tool for cadastral management, territorial planning and decision-making in urban public policies.