Abstract. This study carried out a spatial segmentation of the urban housing market based on a combination of sophisticated data-driven techniques. Such delineation is essential for enhancing the accuracy of mass appraisal and guiding effective urban planning policies. The residential real estate market of Minsk serves as the case study. The analysis relied on a dataset of approximately 4,600 offer prices for secondary market apartments in 2017, sourced from the real estate platform Realt.by. A multi-stage methodological workflow was proposed. An initial evaluation using the Global Moran’s I index (I = 0.39) and Local Indicators of Spatial Association (LISA) confirmed significant price clustering. To address dimensionality and multicollinearity among spatial predictors, Spatial Principal Component Analysis (sPCA) was employed, reducing ten infrastructure variables to four interpretable latent components representing centrality, environmental quality, social-industrial balance, and transport accessibility. Subsequently, these derived spatial factors as well as structural property attributes (such as area, floor level, and room count) were used as inputs for a Geographically Weighted Regression (GWR) model. This specification demonstrated substantial explanatory power (R² = 0.58) and successfully accounted for spatial heterogeneity, eliminating residual autocorrelation. Finally, the local GWR coefficients were grouped using k-means clustering, delineating three distinct submarkets with unique pricing mechanisms: a Central Urbanized zone, driven primarily by factors such as centrality and the number of floors; a Developed Middle-Ring submarket, influenced mainly by property attributes including room count and construction year; and a Modern Peripheral submarket shaped strongly by construction year and the “Centrality and Prestige” component.
Alieva et al. (Sat,) studied this question.