Mapping heterogeneity: Spatially explicit machine learning approaches for urban value uplift characterisation and prediction

Abstract

Understanding urban value uplift is essential for guiding public efforts towards vibrant economies and inclusive communities. Current research, however, often overlooks non-economic dimensions of urban values, resulting in biased socioeconomic outcomes and neglected socio-spatial heterogeneity for sustainable urban planning. This study employs spatially explicit machine learning (ML) approaches to comprehensively investigate the characterisation of urban value uplift and its predictors. Applying GWPCA to the municipality of Shanghai, the results reveal significant heterogeneity in urban value uplift characterisation. While the municipality shows a generally homogeneous polycentric growth pattern, central areas exhibit economic growth, and peripheral areas are enhanced with place value. Additionally, combining a Spatially Enhanced CatBoost algorithm (CTB-S) and SHAPley value, analysis on uplift predictors indicates that urban policy, among many other features, tends to catalyse further economic uplift specifically in central areas. Our findings underscore the inherent heterogeneity in urban change process, necessitating spatially tailored approaches for policymakers to capture urban value uplifts. By addressing the unique needs of different urban areas, policymakers can promote sustainable urban environments that ensure equitable economic and social benefits across all communities.