Incorporating neighborhoods with explainable artificial intelligence for modeling fine-scale housing prices

Abstract

The hedonic price model (HPM) has been widely used to investigate the association between neighborhoods and housing prices. Empirical studies of HPM assume that mixed land use, accessibility, and housing structures generate a substantial premium in housing prices and follow linear relationships, with less attention paid to their nonlinear effects. To fill this gap, this study integrates transaction records over 57,842 housing units in Shanghai and explainable artificial intelligence methods to examine the nonlinear effects of public service amenities, private service amenities, and street view on housing prices. We identified the global threshold effects and their ranges, as well as the local explanations for the price forecast of each housing unit. Nonlinear analysis showed that all public service amenities and some private service amenities (e.g., entertainment) are positively related to housing prices within a certain range, while shopping and carting services are negatively related. Furthermore, the percentage of green pixels in street view images presented a nearly linear relation to housing prices. Residents in Shanghai have paid a premium of about 2000 yuan/m2 for a higher green view. This study contributes to a better understanding of the relationships between housing units and their neighborhoods as well as guidance for the scientific and reasonable formulation of housing prices.