| Title |
Cluster Validity Assessment for Energy Benchmarking Models in Multi-Family Residential Buildings |
| Authors |
문선유(Moon, Sun-Yu) ; 김지형(Kim, Ji-Hyeoung) ; 김의종(Kim, Eui-Jong) |
| DOI |
https://doi.org/10.5659/JAIK.2026.42.10.337 |
| Keywords |
Apartment Complex; Energy Benchmarking; Cluster Validity; K-means Clustering; Regulatory-based Cluster |
| Abstract |
This study evaluates the suitability of regulatory-based clustering criteria for apartment energy benchmarking in South Korea by comparing
cluster validity and benchmarking model performance between regulatory-based and data-driven clusters. The analysis was conducted on 7,347
apartment complexes nationwide using the Building Register and K-apt. The regulatory-based cluster was classified into four groups based on
household size per the Housing Construction Standards Regulations, while the data-driven cluster was constructed by performing K-means
clustering with household size and EUI as input variables and extracting household size boundaries from the clustering results. When the
number of groups was set equally to four, cluster validity showed no significant difference between the two clusters. However, resetting the
optimal number of groups to two resulted in higher cluster validity and benchmarking model performance compared to the regulatory-based
cluster. These findings suggest that reducing the number of groups and incorporating EUI into the clustering criteria can enhance
within-cluster homogeneity, while the current regulatory household boundaries retain a certain level of suitability for energy benchmarking
under equivalent group number conditions. |