エネルギー・資源学会論文誌
Online ISSN : 2433-0531
ISSN-L : 2433-0531
研究論文
家庭CO2統計の個票データと機械学習を用いた建築時期別光熱費の実態把握
西尾 健一郎
著者情報
ジャーナル フリー

2021 年 42 巻 3 号 p. 175-184

詳細
抄録

This study analyzed the factors causing the difference in the amount of utility bill payments by housing construction period, using micro data of about 9,000 households. The analysis consisted of (1) building a model of gradient boosting decision tree, which is one of the machine learning techniques, (2) identifying contribution of each household and each feature using SHAP value, which is a novel method to improve the interpretability of machine learning, and (3) revealing the breakdown of differences on a macro level by aggregating each contribution. The results showed that the factors that have reduced the amount of payments in recently constructed houses were partly caused by the spread of heat pump water heaters and IH cooking heaters, in addition to the improvement of housing insulation performance. It was also confirmed that the effects of higher efficiency refrigerators and lighting have been steadily increasing.

著者関連情報
© 2021 一般社団法人エネルギー・資源学会
前の記事
feedback
Top