Predictive Modeling of Housing Anxiety Using Explainable Artificial Intelligence: Focusing on Single-Person Households in the Seoul Metropolitan Area, Korea

Korea has witnessed a significant increase in the number of vulnerable single-person households facing housing instability, particularly in the Seoul Metropolitan Area. Hence, identifying potential policy targets and implementing preventive measures before such households experience severe instability are essential. In this study, we utilized explainable artificial intelligence—specifically, ensemble-based machine learning techniques (e.g., bagging, random forests, boosting) that use multiple trees to enhance predictive performance—to develop a predictive model for housing instability. This model outperformed traditional classification models (e.g., logistic regression) in predictive accuracy and can help streamline the identification process for households at risk of housing crises. Further, it is valuable for resource-constrained public organizations.