Determinants of Jeonse Prices in Gangnam District: Machine Learning and Explainable Artificial Intelligence Approach

Abstract

This study employs a machine learning model with a black box nature to explore the variables influencing Jeonse apartment prices in the Gangnam district. While traditional real estate finance has relied on Linear Regression, recent advancements have been made by sophisticated machine learning models such as XGBoost, leading to heightened performance. Nevertheless, the inherent opacity of XGBoost poses challenges in comprehending the factors guiding Jeonse prices. Addressing this limitation, we apply TreeSHAP, an eXplainable Artificial Intelligence (XAI) technique, to the XGBoost model, thus elucidating its contributions and facilitating an in-depth analysis of the determinants governing Jeonse prices in the Gangnam district. Our experiments confirm that XGBoost achieves superior performance, compared to linear regression. We delve into influential determinants such as the construction date, major construction companies, and transportation convenience via XAI. This study demonstrates that enhancing the reliability and usability of machine learning can improve the explanatory power of the Jeonse real estate market and its price determinants.