A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Determinants of Electricity Consumption of Energy-Vulnerable Group Using Ensemble Gradient-Boosting Algorithm
https://orcid.org/http://orcid.org/0000-0002-1736-8747
https://orcid.org/http://orcid.org/0000-0001-5774-6688
https://orcid.org/http://orcid.org/0000-0001-7712-4237
The increasing energy burden on vulnerable households is critical in modern cities, it is crucial to understand how cities can characterize energy vulnerability and its relationship with the environment. This study modeled relationships between energy consumption and built environmental factors to compare determinants in average and energy-vulnerable households. While the conventional approach of identifying energy vulnerability often relies on household income, this study suggested a new approach by considering the energy-vulnerable group as a low-income class with high energy expenditure. A traditional regression model (semi-log regression) and advanced machine learning algorithm (ensemble gradient boosting, XGboost) were employed to maximize the performance of the modeling processes. The results indicated that the overall modeling performance was superior with regard to the machine learning algorithm, producing the r-squared value of 0.92 for the energy-vulnerable households, compared to the 0.34 of the semi-log regression model. While the direction of the association of the determinants was similar in the average and energy-vulnerable households, the level of association exhibited a clear difference, especially for the effect of income (comparing 0.30 to 0.03) and housing type (comparing -0.45 to -0.63). The study identified several implications regarding urban energy management and policy based on the findings.
Determinants of Electricity Consumption of Energy-Vulnerable Group Using Ensemble Gradient-Boosting Algorithm
https://orcid.org/http://orcid.org/0000-0002-1736-8747
https://orcid.org/http://orcid.org/0000-0001-5774-6688
https://orcid.org/http://orcid.org/0000-0001-7712-4237
The increasing energy burden on vulnerable households is critical in modern cities, it is crucial to understand how cities can characterize energy vulnerability and its relationship with the environment. This study modeled relationships between energy consumption and built environmental factors to compare determinants in average and energy-vulnerable households. While the conventional approach of identifying energy vulnerability often relies on household income, this study suggested a new approach by considering the energy-vulnerable group as a low-income class with high energy expenditure. A traditional regression model (semi-log regression) and advanced machine learning algorithm (ensemble gradient boosting, XGboost) were employed to maximize the performance of the modeling processes. The results indicated that the overall modeling performance was superior with regard to the machine learning algorithm, producing the r-squared value of 0.92 for the energy-vulnerable households, compared to the 0.34 of the semi-log regression model. While the direction of the association of the determinants was similar in the average and energy-vulnerable households, the level of association exhibited a clear difference, especially for the effect of income (comparing 0.30 to 0.03) and housing type (comparing -0.45 to -0.63). The study identified several implications regarding urban energy management and policy based on the findings.
Determinants of Electricity Consumption of Energy-Vulnerable Group Using Ensemble Gradient-Boosting Algorithm
https://orcid.org/http://orcid.org/0000-0002-1736-8747
https://orcid.org/http://orcid.org/0000-0001-5774-6688
https://orcid.org/http://orcid.org/0000-0001-7712-4237
The increasing energy burden on vulnerable households is critical in modern cities, it is crucial to understand how cities can characterize energy vulnerability and its relationship with the environment. This study modeled relationships between energy consumption and built environmental factors to compare determinants in average and energy-vulnerable households. While the conventional approach of identifying energy vulnerability often relies on household income, this study suggested a new approach by considering the energy-vulnerable group as a low-income class with high energy expenditure. A traditional regression model (semi-log regression) and advanced machine learning algorithm (ensemble gradient boosting, XGboost) were employed to maximize the performance of the modeling processes. The results indicated that the overall modeling performance was superior with regard to the machine learning algorithm, producing the r-squared value of 0.92 for the energy-vulnerable households, compared to the 0.34 of the semi-log regression model. While the direction of the association of the determinants was similar in the average and energy-vulnerable households, the level of association exhibited a clear difference, especially for the effect of income (comparing 0.30 to 0.03) and housing type (comparing -0.45 to -0.63). The study identified several implications regarding urban energy management and policy based on the findings.
Determinants of Electricity Consumption of Energy-Vulnerable Group Using Ensemble Gradient-Boosting Algorithm
KSCE J Civ Eng
Kim, Hyunsoo (author) / Kwon, Youngwoo (author) / Choi, Yeol (author)
KSCE Journal of Civil Engineering ; 26 ; 5010-5021
2022-12-01
12 pages
Article (Journal)
Electronic Resource
English
Heterogeneity of electricity consumption patterns in vulnerable households
| BASE | 2020
Prediction of Home Energy Consumption Based on Gradient Boosting Regression Tree
| BASE | 2021
Determinants of electricity consumption in Dutch dwellings
| Online Contents | 2013