Carbon price prediction in China based on ensemble empirical mode decomposition and machine learning algorithms

Area of Science:

• Environmental Economics
• Machine Learning
• Financial Forecasting

Background:

• Carbon emission trading markets are essential for environmental protection and energy conservation.
• Accurate carbon price forecasting supports China's participation in international carbon finance and pilot market development.
• Carbon price signals are often non-stationary, posing challenges for traditional forecasting methods.

Purpose of the Study:

• To develop and evaluate hybrid machine learning models for accurate short-term carbon price forecasting.
• To enhance the prediction accuracy of carbon prices in China's pilot markets.
• To explore advanced decomposition and optimization techniques for non-stationary financial time series.

Main Methods:

• Ensemble Empirical Mode Decomposition (EEDM) was used to decompose non-stationary carbon price data into intrinsic mode functions (IMFs) and residuals.
• A hybrid Genetic Algorithm (GA) and Back Propagation (BP) neural network model was developed for short-term price prediction, overcoming local optimization issues.
• A hybrid Least Squares Support Vector Machine (LSSVM) and Particle Swarm Optimization (PSO) model was employed to minimize forecast error and search parameters.

Main Results:

• The hybrid GA-BP model demonstrated superior performance for short-term carbon price prediction compared to other algorithms.
• The hybrid PSO-LSSVM model effectively reduced forecast errors and optimized search parameters, outperforming traditional neural network approaches.
• Empirical analysis focused on Guangdong, Hubei, and Shenzhen, key pilot carbon markets in China.

Conclusions:

• Hybrid machine learning models integrating decomposition and optimization techniques offer significant improvements in carbon price forecasting accuracy.
• The proposed methods provide valuable tools for policymakers and market participants in China's developing carbon markets.
• Advanced forecasting techniques are crucial for effective carbon emission trading and environmental policy implementation.