Proposal for Reducing the Burden of Field Surveys of Vacant Houses Using Municipal Owned Data and Machine Learning to Identify Spatial Distribution of Non-Vacant Houses

Abstract

This study is conducted to reduce the burden of field survey work on vacant houses, which currently relies on visual surveys conducted by local governments. First, for the entire area of Maebashi city in Gunma prefecture, a machine learning model was constructed to estimate the probability of vacant houses for each detached house by utilizing the municipality owned data, which provides information on the residents of each building and water consumption. Next, we developed a method to identify detached houses that do not require on-site surveys based on the estimated probability of vacant houses for each detached house using this model. As a result, when detached houses with an estimated vacancy rate of 30% or less were estimated as non-vacant house, we were able to give a determination of non-vacant house to 79.74% of the detached house in the city. In addition, of the detached house determined to be non-vacant house, 99.02% of the buildings were truly non-vacant house. We also found that the districts with many buildings have a higher estimated number of detached houses.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 2 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.