2020 Volume 41 Issue 6 Pages 328-335
Attention has been focused on efforts to promote energy saving actions by utilizing various data regarding energy use of households. When promoting energy-saving behavior, analysis and classification of households’ equipment ownership status will be able to be used for the personalization of energy-saving tips such as energy-saving measures and home appliance replacement in order to promote behavior change effectively. In this paper, we compared the accuracy of methods for classifying equipment ownership status using the national survey of carbon dioxide emissions from residential sector. We employed four classification methods for comparison: binary logistic regression, decision tree, random forest and XGBoosting. We also evaluated the importance of explanatory variables by using permutation test, which remove variables’ feature by randomizing the values of each variable. We find that machine learning methods such as Random Forest and XGBoosting generate relatively higher classification accuracy. Furthermore, these methods show less degradation of classification accuracy when important explanatory variables are permuted from the full model.