技术与创新管理2026，Vol.47Issue(2)：136-149,14.DOI:10.14090/j.cnki.jscx.2026.0202
基于可解释深度学习的中国煤炭价格驱动因素研究
The Study on the Driving Factors of China's Coal Prices Based on Explainable Deep Learning
摘要
Abstract
With the deepening transformation of China's energy structure,fluctuations in coal prices are increasingly influenced by the complex interactions of multiple nonlinear factors.Coal price volatility not only directly affects national energy security but also exerts a profound impact on macroeconomic operations and policy regulation.Therefore,systematic research is urgently needed.This study con-structs a GS-XGBoost-SHAP-based analytical framework to systematically identify the key driving fac-tors of coal price fluctuations from a nonlinear perspective and to uncover the nonlinear mechanisms underlying both individual and interactive variable effects.The results indicate that:the NEWC Austral-ian thermal coal price,Daqing crude oil price,BRENT crude oil price,economic growth,and economic policy uncertainty are the core variables influencing coal price volatility,confirming the"energy-econo-my-uncertainty"three-dimensional driving mechanism of price formation.The relationships between coal prices and key variables exhibit significant nonlinear and asymmetric characteristics,with positive driving effects being notably stronger than negative inhibitory effects.The interaction effects among variables are heterogeneous and display strong nonlinearity in coal price formation,with strong interactions primarily concentrated in the value ranges where key variables exert strong positive effects on coal prices.关键词
中国煤炭价格/非线性影响/GS-XGBoost模型/SHAP可解释性分析/交互作用Key words
coal price in China/nonlinear effects/GS-XGBoost model/SHAP interpretability analysis/interaction effects分类
管理科学引用本文复制引用
吕靖烨,李冲,樊秀峰..基于可解释深度学习的中国煤炭价格驱动因素研究[J].技术与创新管理,2026,47(2):136-149,14.基金项目
教育部人文社会科学研究规划基金项目"气候不确定性对碳排放权市场时变溢出效应测度与政策优化研究"(24YJA790041) （24YJA790041）
