| Title |
Analysis of Factor Influencing Vacant House Occurrence in Depopulation Regions and Development of an Estimation Model using Machine Learning: A Case Study of Gongju-si |
| Authors |
정연준(Jeong, Yeon-Jun) ; 이경환(Lee, Kyung-Hwan) |
| DOI |
https://doi.org/10.38195/judik.2024.02.25.4.39 |
| Keywords |
빈집; 머신러닝; 추정모델; 인구감소지역 Vacant Houses; Machine Learning; Prediction Model; Depopulation Regions |
| Abstract |
The purpose of this study is to derive factors influencing the occurrence of vacant houses and develop a vacant houses prediction model using machine learning to improve the efficiency of the vacant house estimation method in depopulation region. To summarize the research results, first, in a review of previous research on factors affecting the occurrence of vacant houses, it was confirmed that vacant houses occur due to a complex effect of influencing factors of various scales and fields, which were divided into 20 influencing factors in Three sectors. Second, as a result of learning three models by setting the vacant house impact factors derived through previous research review and the confirmed vacant house data in Gongju City as learning data and target data, the XGBoost model showed the highest performance. In addition, it was confirmed that social and economic factors were also factors of high importance and that it was necessary to estimate vacant homes based on complex data. Third, when we compared the prediction results with the spatial environmental characteristics of Gongju City, it was predicted that areas with dense multi-household housing in the new downtown were likely to have vacant houses. In the old downtown and rural areas, it was predicted that areas with old detached house on slopes would have a high possibility of vacant houses. This study suggested a way to streamline the vacant house estimation process by developing a comprehensive data-based vacant house estimation model. |