| Autor: |
Kangalli Uyar, Sinem Guler1 (AUTHOR) skangalli@pau.edu.tr, Dal, Berker2 (AUTHOR) bdal@pau.edu.tr, Ozbay, Bilge Kagan1 (AUTHOR) bkozbay@pau.edu.tr |
| Předmět: |
|
| Zdroj: |
Building & Environment. Aug2024, Vol. 262, pN.PAG-N.PAG. 1p. |
| Abstrakt: |
Controlling carbon emissions is critical to mitigating the negative environmental impacts of carbon emissions, such as global warming and climate change. Identifying the factors that affect building emissions is important because they contribute significantly to global emissions. The study aimed to determine the building characteristics that affect building emissions for residential buildings in Istanbul using the MARS algorithm. The MARS algorithm was also used to estimate building emissions based on the energy performance certificate data for each building. The study revealed that building characteristics have specific thresholds that affect building emissions differently depending on whether they are above or below these thresholds. The feature selection analysis revealed that the thermal insulation properties, specifically the wall_u value, roof_u value, and window_u value, had the greatest impact on building emissions. Other factors identified were building age, heating power, and floor height. These factors alone explain about 77 % of the building emissions. The analysis identified the interactions between building characteristics that affect building emissions. This enables common, appropriate solutions to be developed in terms of building characteristics to reduce building emissions. • We investigate the performance of MARS in estimating emission rates of buildings. • The MARS model has predicted the carbon emission rates of buildings satisfactorily. • The MARS has achieved better prediction accuracy compared to MLR. • The U-values are identified as the critical features impacting on emissions. • Threshold values are determined for building features and examined. [ABSTRACT FROM AUTHOR] |
| Databáze: |
GreenFILE |
| Externí odkaz: |
|
Full Text from ScienceDirect