Scholarlink

<p>Understanding the predictive associations among anthropogenic <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(CO_{2}\)</EquationSource> </InlineEquation> emissions, economic development, and land-use change is important for sustainability-oriented analysis in vulnerable coastal regions. This study examines these environmental-economic linkages in four member countries of the Indian Ocean Rim Association (IORA), namely Malaysia, Mauritius, Sri Lanka, and Madagascar, using historical data from 1960 to 2020. To this end, a machine learning-based forecasting framework, termed Xavier Online Sequential Extreme Learning Machine with Genetic Algorithm (XOS-ELM-GA), is proposed to predict GDP and agricultural land area using <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(CO_{2}\)</EquationSource> </InlineEquation> emissions as a key explanatory indicator. Given the nonlinear and dynamic nature of the statistical relationships among these variables, advanced forecasting methods are required. Preliminary statistical analyses were first conducted to examine the historical relationships among <InlineEquation ID="IEq5"> <EquationSource Format="TEX">\(CO_{2}\)</EquationSource> </InlineEquation> emissions, GDP, and agricultural land area in the selected countries. On this basis, the proposed XOS-ELM-GA framework was developed to perform correlation-based forecasting rather than causal inference. Methodologically, the model integrates Xavier-based weight initialization and an optimized genetic algorithm to improve forecasting accuracy while mitigating common limitations of conventional OS-ELM, including prediction uncertainty, parameter sensitivity, and computational inefficiency. Experimental results show that XOS-ELM-GA consistently outperformed ELM, OS-ELM, and OS-ELM-GA across the four IORA countries. For example, in annual GDP forecasting, it achieved average MSE/SMAPE values of 4.47E-03/10.13% for Malaysia, 1.12E-03/10.48% for Mauritius, and 2.98E-02/12.13% for Sri Lanka. In agricultural land forecasting, the model also delivered competitive performance, with average SMAPE values of 3.77% for Sri Lanka and 2.99% for Madagascar. These findings demonstrate that the proposed model provides more reliable forecasts than standard ELM-based online variants. The resulting forecasts offer useful decision support for policymakers seeking to balance economic development with environmental sustainability in IORA countries, and provide practical insights for advancing the United Nations Sustainable Development Goals, particularly SDG 8, SDG 13, and SDG 15.</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

Machine learning-based forecasting of CO2-related economic growth and agricultural land change in IORA countries

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月9日

查看全文
  • Xiaohan Xu,
  • Guojie Liang,
  • Roy Anthony Rogers,
  • Chyan Zheng Siow

摘要

Understanding the predictive associations among anthropogenic \(CO_{2}\) emissions, economic development, and land-use change is important for sustainability-oriented analysis in vulnerable coastal regions. This study examines these environmental-economic linkages in four member countries of the Indian Ocean Rim Association (IORA), namely Malaysia, Mauritius, Sri Lanka, and Madagascar, using historical data from 1960 to 2020. To this end, a machine learning-based forecasting framework, termed Xavier Online Sequential Extreme Learning Machine with Genetic Algorithm (XOS-ELM-GA), is proposed to predict GDP and agricultural land area using \(CO_{2}\) emissions as a key explanatory indicator. Given the nonlinear and dynamic nature of the statistical relationships among these variables, advanced forecasting methods are required. Preliminary statistical analyses were first conducted to examine the historical relationships among \(CO_{2}\) emissions, GDP, and agricultural land area in the selected countries. On this basis, the proposed XOS-ELM-GA framework was developed to perform correlation-based forecasting rather than causal inference. Methodologically, the model integrates Xavier-based weight initialization and an optimized genetic algorithm to improve forecasting accuracy while mitigating common limitations of conventional OS-ELM, including prediction uncertainty, parameter sensitivity, and computational inefficiency. Experimental results show that XOS-ELM-GA consistently outperformed ELM, OS-ELM, and OS-ELM-GA across the four IORA countries. For example, in annual GDP forecasting, it achieved average MSE/SMAPE values of 4.47E-03/10.13% for Malaysia, 1.12E-03/10.48% for Mauritius, and 2.98E-02/12.13% for Sri Lanka. In agricultural land forecasting, the model also delivered competitive performance, with average SMAPE values of 3.77% for Sri Lanka and 2.99% for Madagascar. These findings demonstrate that the proposed model provides more reliable forecasts than standard ELM-based online variants. The resulting forecasts offer useful decision support for policymakers seeking to balance economic development with environmental sustainability in IORA countries, and provide practical insights for advancing the United Nations Sustainable Development Goals, particularly SDG 8, SDG 13, and SDG 15.