Propuesta metodológica basada en regresión espacial kriging para estimar el valor por metro cuadrado de área privada en apartamentos sometidos a propiedad horizontal con fines de avalúo hipotecario (2023– 2025): Caso UPL Britalia, Bogotá D.C.

Abstract

This project proposes a methodological approach for estimating the square meter value of private area in apartments under condominium ownership, focusing on the Local Planning Unit (UPL) Britalia in Bogotá D.C. during the 2023–2025 period. The proposal arises from the need to overcome the limitations of traditional urban valuation methods, which often rely on direct comparisons or zonal averages that fail to adequately capture spatial heterogeneity and contextual conditions influencing real estate market behavior. To address this, a georeferenced database is constructed by integrating technical appraisal records from the study period with real estate listings systematically collected through web scraping from specialized portals, ensuring a consistent and up-to-date input for analysis. On this basis, regression models and spatial econometrics are applied to evaluate the impact of physical, socioeconomic, and territorial variables on unit values. Additionally, geostatistical techniques, particularly residual kriging, are considered to capture spatial dependence not explained by econometric models, allowing the generation of estimation surfaces and cartographic representations consistent with the territorial distribution of property values. The outcomes of this approach are projected as a replicable methodological tool that broadens the technical alternatives available for urban valuation, supporting the validation of appraisals, the interpretation of spatial market structures, and the strengthening of traceability in real estate management processes. Furthermore, the scalable nature of the methodology opens the possibility of adapting it to other urban contexts characterized by high heterogeneity and partial data availability, provided that georeferenced information is available to support model structuring.