Understanding Housing Prices Using Geographic Big Data: A Case Study in Shenzhen

Abstract

Understanding the spatial pattern of urban house prices and its association with the built environment is of great significance to housing policymaking and urban planning. However, many studies on the influencing factors of urban housing prices conduct qualitative analyses using statistical data and manual survey data. In addition, traditional housing price models are mostly linear models that cannot explain the distribution of housing prices in urban areas. In this paper, we propose using geographic big data and zonal nonlinear feature machine learning models to understand housing prices. First, the housing price influencing factor system is built based on the hedonic pricing model and geographic big data, and it includes commercial development, transportation, infrastructure, location, education, environment, and residents’ consumption level. Second, a spatial exploratory analysis framework for house price data was constructed using Moran’s I tools and geographic detectors. Finally, the XGBoost model is developed to assess the importance of the variables influencing housing prices, and the zonal nonlinear feature model is built to predict housing prices based on spatial exploration results. Taking Shenzhen as an example, this paper explored the distribution law of housing prices, analyzed the influencing factors of housing prices, and compared the different housing price models. The results show that the zonal nonlinear feature model has higher accuracy than the linear model and the global model.