Actualización masiva del componente económico del catastro rural en el municipio de Sasaima (Cundinamarca) para la reducción del rezago de avalúos mediante métodos de aprendizaje automático para el año 2025

Abstract

This study addresses the lag in rural cadastral valuations in the municipality of Sasaima by developing an indirect methodology supported by advanced data analytics for updating the economic component. A dataset of 327 real estate offers from the Gualivá province was integrated, homogenized to the year 2025 using the CPI, and filtered through statistical analysis and Local Indicators of Spatial Association (LISA). The methodology combined a Random Forest model to capture non-linear relationships between biophysical, agrological, and socioeconomic covariates with the Unit Land Value, alongside a geostatistical interpolation of the residuals using Regression-Kriging to account for the spatial stochastic component. The results demonstrated that this hybrid approach outperforms traditional linear estimations and conventional Kriging, offering a precise, scalable tool aligned with the Colombian regulatory framework that strengthens tax equity and Multipurpose Cadastre management.