Hybrid Machine Learning Analysis of Exogenous Features in China’s Guangdong Carbon Market Price Prediction
中国広東炭素市場価格予測における外生変数のハイブリッド機械学習分析 (AI 翻訳)
Chenyao Duan, Yuanfang Chen, Junlin He, Kok-Haur Ng
🤖 gxceed AI 要約
日本語
中国最大の炭素市場である広東省の炭素価格を、CNN+LSTMのハイブリッドモデルで予測。外生変数（原油価格、取引量）の寄与を分析し、ラグ付き日中価格が最も重要であることを示した。
English
This study applies a hybrid CNN-LSTM model to forecast carbon prices in China's largest carbon market, Guangdong. It demonstrates the model's superiority over standalone LSTM and finds that lagged intraday OHLC prices are the most critical predictors, while crude oil prices and trading volume contribute moderately.
Unofficial AI-generated summary based on the public title and abstract. Not an official translation.
📝 gxceed 編集解説 — Why this matters
日本のGX文脈において
日本では2026年度から本格稼働する排出量取引制度（GX-ETS）の参考となる。本論文の機械学習アプローチは、日本の炭素価格予測にも応用可能であり、価格変動リスク管理に役立つ。
In the global GX context
The hybrid ML approach for carbon price prediction offers a methodology applicable to global emissions trading systems (ETS). It highlights the importance of intraday price dynamics and external factors like oil prices, relevant for market efficiency and risk management in carbon markets worldwide.
👥 読者別の含意
🔬研究者:Demonstrates the effectiveness of hybrid deep learning for carbon price forecasting, providing a benchmark for future studies on exogenous feature selection.
🏢実務担当者:Carbon market participants (e.g., compliance buyers, traders) can use the model to improve price prediction and inform hedging strategies.
🏛政策担当者:Insights into key price drivers can inform market design and regulatory oversight to enhance carbon market stability.
📄 Abstract（原文）
Carbon pricing is essential for the effective functioning of carbon markets, providing a basis for strategies that support sustainable green economic development. This study investigates the predictive power of intraday carbon price-related information, trading volume and crude oil prices within a hybrid model combining a convolutional neural network (CNN) and a long short-term memory (LSTM) network to forecast carbon emission allowance closing prices. The model is applied to Guangdong, China’s largest and most active carbon-trading market, to improve forecasting accuracy. The CNNLSTM models are benchmarked against a standalone LSTM model using multiple loss functions, with results showing the hybrid model’s clear superiority in both in-sample and out-of-sample forecasts. Analysis of exogenous variable highlights that lagged intraday opening, highest, lowest and closing (OHLC) prices are the most critical predictors, with their exclusion significantly reducing accuracy. Brent crude oil prices and trading volume provide moderate contributions to the model, while quasi-likelihood analysis reaffirms lagged OHLC information and oil prices as key factors in carbon price prediction.
🔗 Provenance — このレコードを発見したソース
- semanticscholar https://doi.org/10.21315/aamjaf2026.22.1.3first seen 2026-07-04 05:18:28
🔔 こうした論文の新着を逃したくない方は キーワードアラート に登録（無料・3キーワードまで）。
gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。 最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。