DKWM-XLSTM: A Carbon Trading Price Prediction Model Considering Multiple Influencing Factors

Area of Science:

• Environmental Science
• Climate Change Mitigation
• Financial Markets

Background:

• Forestry carbon sinks are vital for climate change mitigation and carbon trading systems.
• Remote sensing is crucial for monitoring carbon sinks and forecasting carbon prices.
• Carbon price prediction is challenging due to non-stationarity and uncertainty.

Purpose of the Study:

• To develop a robust method for predicting multi-factor influenced carbon trading prices.
• To address the complexities of non-stationary data and inherent uncertainties in carbon markets.
• To enhance the accuracy and reliability of carbon price forecasting.

Main Methods:

• A Decomposition (DECOMP) module separates data into trend and cyclical components.
• A KAN with Multi-Domain Diffusion (KAN-MD) module extracts features and manages non-stationarity.
• A Wave-MH attention module, using wavelet transformation, reduces uncertainty interference.

Main Results:

• The proposed model demonstrates superior predictive accuracy in the Hubei carbon trading market.
• Achieved Mean Squared Error (MSE) of 0.204% and Mean Absolute Error (MAE) of 0.0277.
• The model exhibits enhanced resilience to price fluctuations compared to benchmark methods.

Conclusions:

• The developed method offers a reliable approach for carbon price prediction.
• Results support improved pricing of carbon financial derivatives.
• The study contributes to better risk management in carbon trading markets.