Exploring the Nonlinear Relationship between Development Density and Housing Price: Focusing on eXplainable AI (XAI)

Abstract

This study aimed to investigate the association between development density and housing prices in Seoul, utilizing a machine learning driven hedonic price model. The model employed actual transaction data of high-rise condominiums (apartments) spanning from January 2019 to December 2021. The study employed CatBoost algorithm and eXplainable AI techniques (SHAP) to explore the non-linear relationships between floor area ratio(FAR) and housing prices. The findings revealed that overall FAR had a negative relationship with housing prices for both general apartments and mixed-use development. However, the negative impact of FAR was relatively small in large-scale development. Furthermore, the negative impact of FAR may not be significant in some high-end mixed-use development. These findings imply the importance of proposing appropriate site plans to ensure a desirable residential environment when high-density development is necessary.