Modified data-driven framework for housing market segmentation

Housing market segmentation is significant at both the conceptual and empirical levels because it reflects the spatial heterogeneity of housing prices, improves the predictive accuracy of housing prices, and indicates dynamic changes in housing markets. The existing literature offers a popular framework, called the data-driven method, to delineate submarkets based on principal component analysis (PCA) and cluster analysis; however, the traditional framework does not consider spatial heterogeneity and has difficulty balancing the spatial relationships (i.e., distance and topological relationships) and attribute similarities. To address these limitations, this paper proposes a modified data-driven framework for delineating housing submarkets by integrating geographically weighted principal component analysis (GWPCA), a spatial heterogeneity test, a density-based spatial clustering (DBSC) algorithm, and hedonic validation. The modified framework is applied to housing-market segmentation in Shenzhen, China. The results indicate that the modified framework exhibits the best performance in submarket segmentation in Shenzhen. The framework has important implications and high potential for identifying housing submarkets statistically, and it can be generalized and applied to housing markets in other cities. In addition, the visualisation results can be used by appraisers for property valuation and by city planners for facility management and social-equality improvement and balance.