6.2 Comparison with Existing Approaches

Urban land valuation has traditionally been addressed through a variety of methodological approaches, including statistical regression models, hedonic pricing methods, spatial econometric models, geographic information system (GIS)-based analyses, and, more recently, machine learning and deep learning techniques. These approaches have significantly improved the accuracy of land-price estimation and have contributed to a better understanding of spatial variations in land markets.

Despite their methodological differences, most existing approaches share a common objective: to estimate land prices from observed explanatory variables. Land price is generally treated as the primary analytical target, while the underlying mechanisms responsible for the formation and spatial organization of urban value receive comparatively less attention.

The ULVF framework adopts a fundamentally different perspective. Rather than beginning with observed land prices, it starts from the spatial mechanisms that generate urban value. Land prices are subsequently determined as the economic expression of the Urban Land Value Field through the sequential construction of the Value Correlation Matrix (VCM), the Value Relation Matrix (VRM), and the Price Relation Matrix (PRM).

This distinction represents a conceptual shift from price-oriented modelling to value-oriented modelling. Within the ULVF framework, transportation networks, accessibility, socio-economic activities, urban value centers, environmental conditions, land-use functions, and localized spatial characteristics are not merely explanatory variables used to improve predictive accuracy. Instead, they constitute the structural mechanisms responsible for generating and organizing urban value across space.

Another distinguishing characteristic of the ULVF framework is its emphasis on spatial relationships rather than isolated observations. Conventional valuation models generally estimate the price of each parcel independently, even when spatial effects are incorporated through neighborhood variables or spatial autocorrelation models. In contrast