Improving carbon dioxide emission predictions through a hybrid model utilising an advanced sparrow search algorithm

Area of Science:

• Environmental Science
• Climate Science
• Data Science

Background:

• Rising carbon dioxide (CO2) emissions drive global warming and climate change, threatening human development and ecosystems.
• Existing CO2 emission prediction models struggle with scarce data, inherent uncertainty, and data volatility.
• Accurate forecasting is crucial for effective climate change mitigation strategies.

Purpose of the Study:

• To develop a novel hybrid model for accurate CO2 emission prediction, especially with limited data.
• To enhance prediction performance through effective feature selection and model parameter optimization.
• To validate the model's efficacy using Chinese carbon emission data.

Main Methods:

• A stable feature screening method was developed for reliable feature selection.
• A least squares support vector machine (LSSVM) was employed for predicting periodic sequences with limited samples.
• The LSSVM parameters were optimized using an improved Sparrow Search Algorithm (SSA) incorporating Sin chaos mapping, adaptive inertia weights, and Cauchy-Gauss variables.

Main Results:

• The improved SSA demonstrated superior global optimization capability and faster convergence compared to standard methods.
• The proposed hybrid model achieved the highest consistency with actual data, significantly improving prediction accuracy.
• Comparative experiments validated the model's superiority over existing approaches.

Conclusions:

• The novel hybrid model effectively addresses the challenges of CO2 emission prediction with limited and complex data.
• The optimized feature selection and LSSVM with enhanced SSA provide a robust framework for accurate environmental forecasting.
• This approach offers a significant advancement in predicting CO2 emissions for climate change research and policy-making.