Predicting Housing Price in Seoul using Explainable AI (XAI) and Machine Learning

• Hae Jung Chun,
• U Hui Lee,
• Bong Gyou Lee,

Vol. 19, No. 4, April 30, 2025
10.3837/tiis.2025.04.002, Download Paper (Free):

• Housing Price
• Machine Learning
• XGBoost
• XAI
• SHAP (SHapley Additive exPlanations)

Abstract

This study analyzed 61,593 cases of real transaction price of apartments in Seoul from January 1, 2021 to September 30, 2023, with the output variable being set as the sales price per dedicated area and the input variables being set as the contract month, floor, year of construction, number of units, and distance to the subway station, etc. After determining which of the machine learning (ML) models, LGBM, XGBoost, and GBDT, had the best predictive power, the importance of each variable was analyzed using XAI's SHAP (SHapley Additive exPlanations) technique. The R2 value of XGBoost was 0.917, MAE value was 134.971, and RMSE value was 191.325, showing the best predictive power. According to the results of applying the SHAP technique to the XGBoost model, heating method, year of construction, distance to subway station, contract month, number of households, distance to market, distance to middle school, distance to high school, distance to elementary school, number of floors, home network, contract date, and management method have the highest influence and importance on the sales price per dedicated area.

Statistics

Cite this article