Carbon price prediction based on a scaled PCA approach

Area of Science:

• Environmental Economics
• Financial Modeling
• Data Science

Background:

• Carbon price prediction is vital for carbon trading market policy and stability.
• Traditional dimensionality reduction methods have limitations in capturing the complexity of influencing factors.
• Accurate carbon price forecasting is essential for market participants and regulators.

Purpose of the Study:

• To introduce and evaluate a novel dimensionality reduction method, scaled principal component analysis (s-PCA), for enhanced carbon price prediction.
• To assess the predictive performance of s-PCA combined with regression and Long Short-Term Memory (LSTM) models.
• To examine the economic value and investment implications of the s-PCA method in carbon markets.

Main Methods:

• Construction of a comprehensive factor library including technical, financial, and commodity indicators.
• Application of scaled principal component analysis (s-PCA) for dimensionality reduction of influencing factors.
• Utilizing traditional regression and Long Short-Term Memory (LSTM) models for carbon price prediction.
• Evaluation of economic value through investment portfolio construction and performance analysis.

Main Results:

• The s-PCA model demonstrated superior performance compared to competing models in both in-sample and out-of-sample predictions on the Hubei Emissions Exchange data.
• Integration with the LSTM model further enhanced the predictive accuracy of the s-PCA method.
• Investors employing the s-PCA method achieved higher returns and Sharpe ratios than with comparative methods or a buy-and-hold strategy.

Conclusions:

• The scaled principal component analysis (s-PCA) method is effective and robust for predicting carbon prices.
• s-PCA offers significant advantages in improving prediction accuracy and efficiency by addressing the limitations of traditional dimensionality reduction techniques.
• The proposed method provides a valuable tool for market participants seeking to optimize investment strategies and enhance market timing in carbon trading.