Scholarlink

<p>The building sector plays a crucial role in urbanization and economic growth. However, the sector is a major emitter of greenhouse gases and dependent on fossil fuel energy. In 2022, it accounted for 37% of global energy and process-related CO<sub>2</sub> emissions, underscoring its impact on climate change. This study aims to predict greenhouse gas emissions from buildings across six African nations—Nigeria, Algeria, South Africa, Egypt, Ethiopia, and Morocco—using time series data spanning from 1980 to 2023 from the World Bank, with future projections extending from 2024 to 2040. The analysis incorporates three key factor categories: energy consumption, demographic characteristics, and economic features. A group of ensemble machine learning models, including random forest (RF), extreme gradient boosting (XGBoost), categorical boosting (CatBoost), and gradient boosting (GB), and non-ensemble machine learning models, including support vector machine (SVM), Bayesian neural networks (BNNs), K-nearest neighbors (KNN), and multilayer perceptron (MLP), were employed alongside the prophet model for time series forecasting. Models were evaluated using tenfold cross-validation for ensemble and non-ensemble models and fourfold cross-validation for prophets, with their performance assessed using R<sup>2</sup>, MAE, MAPE, and rRMSE. Among the models, GB (R<sup>2</sup> = 0.952) and MLP (R<sup>2</sup> = 0.966) delivered the highest predictive accuracy. The ensemble framework achieved an overall accuracy of 0.934. SHAP values were used for model interpretability, revealing total energy consumption as the most significant factor in building-related emissions. National comparisons revealed Nigeria as the most influential country in prediction. In future projections of individual nations’ datasets, Prophet revealed various results among nations, with Morocco emerging as the best accuracy with R<sup>2</sup> of 99.6%. These findings provide valuable insights for African policymakers, emphasizing the need for sustainable urban planning and energy strategies to reduce emissions and align with the United Nations Sustainable Development Goals (SDGs).</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Exploring machine learning models for predicting greenhouse gas emissions in Africa’s building sector: A case study of six nations

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月6日

查看全文
  • Abdullahi Mohamud Adam,
  • Abdikarim Nor Osman,
  • Abdinasir Mohamed Yusuf,
  • Huseyin Gokcekus,
  • Farhad Bolouri

摘要

The building sector plays a crucial role in urbanization and economic growth. However, the sector is a major emitter of greenhouse gases and dependent on fossil fuel energy. In 2022, it accounted for 37% of global energy and process-related CO2 emissions, underscoring its impact on climate change. This study aims to predict greenhouse gas emissions from buildings across six African nations—Nigeria, Algeria, South Africa, Egypt, Ethiopia, and Morocco—using time series data spanning from 1980 to 2023 from the World Bank, with future projections extending from 2024 to 2040. The analysis incorporates three key factor categories: energy consumption, demographic characteristics, and economic features. A group of ensemble machine learning models, including random forest (RF), extreme gradient boosting (XGBoost), categorical boosting (CatBoost), and gradient boosting (GB), and non-ensemble machine learning models, including support vector machine (SVM), Bayesian neural networks (BNNs), K-nearest neighbors (KNN), and multilayer perceptron (MLP), were employed alongside the prophet model for time series forecasting. Models were evaluated using tenfold cross-validation for ensemble and non-ensemble models and fourfold cross-validation for prophets, with their performance assessed using R2, MAE, MAPE, and rRMSE. Among the models, GB (R2 = 0.952) and MLP (R2 = 0.966) delivered the highest predictive accuracy. The ensemble framework achieved an overall accuracy of 0.934. SHAP values were used for model interpretability, revealing total energy consumption as the most significant factor in building-related emissions. National comparisons revealed Nigeria as the most influential country in prediction. In future projections of individual nations’ datasets, Prophet revealed various results among nations, with Morocco emerging as the best accuracy with R2 of 99.6%. These findings provide valuable insights for African policymakers, emphasizing the need for sustainable urban planning and energy strategies to reduce emissions and align with the United Nations Sustainable Development Goals (SDGs).