Carbon price prediction based on modified wavelet least square support vector machine.

Area of Science:

• Environmental Economics
• Computational Finance
• Time Series Analysis

Background:

• Accurate carbon price prediction is crucial for global warming mitigation and China's carbon trading market development.
• High non-determinacy and non-linearity in carbon price series challenge traditional single-model forecasting.
• Existing models struggle to meet the required prediction accuracy for carbon pricing.

Purpose of the Study:

• To develop a novel hybrid forecasting model for enhanced carbon price prediction accuracy.
• To introduce and apply the wavelet least square support vector machine (wLSSVM) in carbon price forecasting.
• To validate the proposed model's effectiveness against multiple benchmark models.

Main Methods:

• Ensemble Empirical Mode Decomposition (EEMD) to decompose carbon price series into stable sub-sequences.
• Linearly Decreasing Weight Particle Swarm Optimization (LDWPSO) to optimize model parameters.
• Wavelet Least Square Support Vector Machine (wLSSVM) for forecasting individual sub-sequences, combined for final prediction.

Main Results:

• The proposed EEMD-LDWPSO-wLSSVM model demonstrated superior forecasting performance compared to 12 other models across three regions.
• The hybrid model significantly improved the accuracy of carbon price prediction.
• Evaluation metrics including MAPE, RMSE, and R2 confirmed the model's effectiveness.

Conclusions:

• The novel hybrid model offers a substantial improvement in carbon price prediction accuracy.
• The EEMD-LDWPSO-wLSSVM approach is effective in handling the complexities of carbon price series.
• This model holds significant potential for widespread application in carbon price forecasting and trading markets.