Optimizing housing price estimation through image segmentation and geographically weighted regression: an empirical study in Nanjing, China

Abstract

Although well-designed urban streets are beneficial for sustainability and livability, few studies have considered their role in housing price estimates. To fill this gap, this study conducted in Nanjing, China, aims to examine the contribution of streetscape features to housing prices. Data were collected for 2040 residential blocks within the four municipal districts in July 2021. A semantic segmentation approach was used to identify the percentage of elements in the images from Baidu Street View. Two types of streetscape related variables (Enclosure and Greenery) were calculated and added to a hedonic pricing model based on Geographically Weighted Regression. The results show that the streetscape factors all have positive effects on house prices, and the contribution to house prices from large to small is grass, plants, horizontal buildings, vertical buildings and trees. By comparing the parameters of the models, it can be concluded that the inclusion of streetscape features and consideration of spatial heterogeneity can significantly improve the accuracy of housing price estimation. The findings of the current study contribute to decision-making in housing planning and urban design and to judgments about pricing reasonableness.