gxceed
← 論文一覧に戻る

Research on Prediction of Carbon Emission Rights Trading Prices of Listed Manufacturing Enterprises in China Based on Ensemble Learning

アンサンブル学習に基づく中国上場製造企業の炭素排出権取引価格予測に関する研究 (AI 翻訳)

Weijie Liu

Proceedings of the 2026 5th International Conference on Big Data, Information and Computer Network学会2026-01-09#炭素価格Origin: CN
DOI: 10.1145/3801228.3801261
原典: https://doi.org/10.1145/3801228.3801261

🤖 gxceed AI 要約

日本語

本論文は、中国の上場製造企業を対象に、アンサンブル学習（LightGBM、XGBoost、CatBoost）を用いて炭素排出権取引価格を予測する枠組みを構築した。SHAP手法によりモデルの解釈性を高め、CatBoostが地域や業種などのカテゴリカル特徴量を含む混合データに最も優れることを示した。取引要因が短期的な炭素価格変動の主要因であることを明らかにし、製造企業の炭素資産管理戦略への示唆を提供する。

English

This paper constructs an ensemble learning prediction framework using LightGBM, XGBoost, and CatBoost to predict carbon emission trading prices for listed manufacturing enterprises in China. Using SHAP for model interpretability, it finds that CatBoost performs best on mixed data with categorical features like region and industry. Trading factors are identified as the core driver of short-term carbon price fluctuations, offering empirical support for differentiated carbon asset management strategies.

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

日本の炭素市場（東京都排出量取引制度、埼玉県目標設定型排出量取引制度など）でも同様のアンサンブル学習手法が応用可能であり、特に製造業に特化した価格予測モデルの知見は、日本企業のカーボンプライシング対応や炭素資産管理に示唆を与える。

In the global GX context

This research contributes to the global literature on carbon price prediction by applying ensemble learning and SHAP to a manufacturing-focused dataset. It provides a replicable methodology for analyzing carbon markets in other regions, particularly for identifying the role of trading factors in short-term price dynamics.

👥 読者別の含意

🔬研究者:Provides empirical evidence on the effectiveness of ensemble learning (especially CatBoost) for carbon price prediction in a manufacturing context, with SHAP-based interpretability.

🏢実務担当者:Manufacturing firms can adopt the modeling framework to forecast carbon costs and optimize carbon asset management, reducing compliance risk.

🏛政策担当者:Offers insights into how trading factors drive carbon price volatility, informing market design and stability mechanisms.

📄 Abstract（原文）

Accurate prediction of carbon emission trading prices is of great significance for enterprises to reduce compliance costs and optimize the allocation of green assets. In view of the nonlinearity, high noise and strong policy-driven characteristics of China's carbon market, this paper constructs an ensemble learning prediction framework based on LightGBM, XGBoost and CatBoost, and uses the SHAP method to open the model “black box”. The study found that CatBoost performs best when dealing with mixed data containing categorical features such as regions and industries. Based on the predictions, this paper reveals that trading factors are the core force driving short-term fluctuations in carbon prices and provides empirical evidence for manufacturing enterprises to develop differentiated carbon asset management strategies.

🔗 Provenance — このレコードを発見したソース

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。