Block-scale modeling of residential land prices: Incorporating multilevel determinants and explainable artificial intelligence

Abstract

Accurate urban residential land price modeling is essential to optimize land allocation, shape tax policies, and promote sustainable urban development. Traditional statistical models often struggle to capture the interactive and nonlinear effects of determinants on land prices. This study integrates multilevel determinants based on hedonic pricing theory with explainable artificial intelligence (XAI) techniques to improve land price modeling. Focusing on Wuhan, we used geographic big data and street view images to evaluate location, neighborhood, and environmental factors at various spatial levels, including nearest accessibility, 15-min walk availability, and residential cluster availability. These factors were incorporated into three tree-based machine learning algorithms, random forest, gradient boosting tree, and eXtreme gradient boosting (XGB), to build predictive models. The XGB model outperformed the others and was used to predict prices in unobserved blocks. SHapley Additive exPlanations were applied to interpret the results, revealing key determinants of land prices. The proximity to rivers and central business districts emerged as significant factors. The influence of urban amenities varied on spatial scales, and green spaces had a stronger impact on the 15-min walk scale than on the larger residential cluster scale. Nonlinear threshold effects were identified, such as the diminishing negative impact of distance to the nearest metro station within a 1.5 km radius, beyond which the effect becomes negligible. Notably, significant interactive effects were observed, particularly the synergistic relationship between riverside locations and the visual presence of street green spaces, which together enhance land value. This study combines hedonic pricing with XAI to improve both predictive accuracy and interpretability, supporting evidence-based decision-making for smart urban planning and governance.