Scholarlink

<p>The research adopts machine learning techniques to forecast carbon prices for carbon trading markets. The forecast accuracy suffers from the complex market dynamics of carbon pricing together with legislative changes in the system. The extraction of features combined a multi-layer perceptron (MLP) neural network model with Random Forest (RF), AdaBoost and CatBoost, LightGBM, XGBoost and HGBoost as machine learning methods. Wavelet decomposition along with autocorrelation functions served as the tools for analyzing time series data. The analysis revealed both immediate and further future dependencies in the data. Three different performance metrics - R2, RMSE and MAE - provided datasets for determining model reliability. The MLP-CatBoost model reached its highest predictive accuracy level during testing which was confirmed through an R2 value of 0.9658. During training the MLP-RF model demonstrated excellent performance because its R2 value reached 0.9967. The results verify that machine learning methods should be used for predicting carbon prices. These findings also present knowledge that researchers can use to develop better models and apply them practically.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Utilising cutting-edge forecasting methods to investigate the unstable environmental conditions of carbon markets: highlighting potential futures in the economic and environmental domains

  • 期刊论文

  • 发表日期：

    2025年12月26日

查看全文
  • Chanjuan Li

摘要

The research adopts machine learning techniques to forecast carbon prices for carbon trading markets. The forecast accuracy suffers from the complex market dynamics of carbon pricing together with legislative changes in the system. The extraction of features combined a multi-layer perceptron (MLP) neural network model with Random Forest (RF), AdaBoost and CatBoost, LightGBM, XGBoost and HGBoost as machine learning methods. Wavelet decomposition along with autocorrelation functions served as the tools for analyzing time series data. The analysis revealed both immediate and further future dependencies in the data. Three different performance metrics - R2, RMSE and MAE - provided datasets for determining model reliability. The MLP-CatBoost model reached its highest predictive accuracy level during testing which was confirmed through an R2 value of 0.9658. During training the MLP-RF model demonstrated excellent performance because its R2 value reached 0.9967. The results verify that machine learning methods should be used for predicting carbon prices. These findings also present knowledge that researchers can use to develop better models and apply them practically.