This study aims to develop a machine learning-based model to predict the risk of Jeonse deposit guarantee accidents by integrating accident data from the Housing and Urban Guarantee Corporation (HUG) with various real estate market indicators. A regio...
This study aims to develop a machine learning-based model to predict the risk of Jeonse deposit guarantee accidents by integrating accident data from the Housing and Urban Guarantee Corporation (HUG) with various real estate market indicators. A regional monthly panel dataset was constructed by combining HUG’s guarantee issuance and accident records with multiple public data sources, including transaction information from the Ministry of Land, Infrastructure and Transport, housing price indices from the Korea Real Estate Board, demographic and household statistics from Statistics Korea, and interest rate data from the Bank of Korea. The Random Forest algorithm was applied to this integrated dataset to estimate accident occurrence probabilities. The model achieved strong predictive performance, with an accuracy of 0.97, recall of 0.97, precision of 1.00, F1-score of 0.98, and ROC-AUC of 0.93. Feature importance analysis revealed that accident rate, Jeonse price index change, base rate fluctuation, new guarantee issuance volume, and population growth rate were major influencing variables. The findings demonstrate the empirical potential of public data integration for predictive risk management and are expected to contribute to enhancing HUG’s Jeonse guarantee risk management and early warning systems.