When will China's industrial carbon emissions peak? Evidence from machine learning

Area of Science:

• Environmental Science
• Data Science
• Industrial Ecology

Background:

• Industrial manufacturing is a primary contributor to global carbon emissions.
• Accurate prediction of industrial carbon emissions is vital for climate change mitigation and low-carbon economic strategies.
• Effective control measures necessitate reliable forecasting models.

Purpose of the Study:

• To compare the predictive accuracy of Support Vector Machine (SVM) and Backpropagation (BP) neural networks for industrial carbon emissions.
• To identify the most effective machine learning model for forecasting industrial CO2 emissions from 2021 to 2040.
• To utilize feature selection to improve model performance.

Main Methods:

• Utilized a random forest technique for feature selection to reduce noise in input variables.
• Employed Support Vector Machine (SVM) and Backpropagation (BP) neural networks for predictive modeling.
• Compared the predictive performance of SVM and BP models using historical industrial carbon emission data.

Main Results:

• BP neural networks demonstrated poor adaptability to small sample sizes, high error rates, and inconsistent predictions.
• SVM models achieved excellent fitting results with small datasets, providing highly accurate predictions.
• Industrial carbon emissions are projected to peak in the year 2030.

Conclusions:

• Support Vector Machines are superior to BP neural networks for predicting industrial carbon emissions, especially with limited data.
• Accurate forecasting models are essential for developing effective industrial emission control policies.
• The study predicts a peak in industrial carbon emissions by 2030, highlighting the urgency for mitigation efforts.