no113-117

[115] Analysis of Occurrence Factors of Vacant Houses by Interpretability Methods in Machine Learning

Aiming to Development of Tools that Support Measures to Solve Vacant House Problems, which is the Best Fit for the Place

○Katsuya Mizusawa¹, Kei Miyamoto², Shota Tamura³, Takahiro Tanaka³ (1. CTI Engineering Co., Ltd., 2. Pasco Corporation, 3. Hiroshima University)

Keywords:Vacant House, Prediction Model, Machine Leaning, Interpretability Methods

In recent years, the declining population and the corresponding increase in vacant houses has become an issue, and efforts to prevent the occurrence of vacant houses are needed. Thus, areal measures are needed against the vacancy of houses in the future for urban planning. Therefore, in this study, we aimed to develop a “vacant house countermeasure support tool” that can spatially identify “which buildings in which areas are likely to become vacant,” and proposed a method for analysing occurrence factor of vacancy of houses using machine learning interpretation techniques. As a result, it became clear that that the performance of buildings and the condition of infrastructure affect the occurrence of vacant houses and that the factors that cause the occurrence of vacant houses differ according to the characteristics of each district, and the degree to which these factors affect the occurrence of vacant houses also differs.