| 基于机器学习的中国铁矿石资源产业链评估 |
| 点此下载全文 |
| 引用本文：李林泰,江飞涛,李海玲,谢聪敏,张艳飞,邵留国.2025.基于机器学习的中国铁矿石资源产业链评估[J].地球学报,46(5):991-1006. |
| DOI：10.3975/cagsb.2025.082703 |
| 摘要点击次数: 308 |
| 全文下载次数: 257 |
|
| 基金项目:国家科技重大专项(编号: 2024ZD1002002);国家自然科学基金项目(编号: 72373160; 72203059);教育部人文社科项目(编号: 22YJCZH078);国家社科基金重大项目(编号: 22&ZD098) |
|
| 中文摘要:由于国内铁矿石资源品位低, 且供应量不足, 中国炼铁生产对进口铁矿石依赖性较强, 因此科学评估中国铁矿石资源保障情况显得尤为重要。本文旨在定量评估中国铁矿石资源的供应安全, 据此进行预警, 并针对现有风险问题提出相应的策略与措施。本文一是构建涵盖铁矿石保障因素、经济发展因素、运输周转因素、经济风险因素、价格波动因素的评价体系, 确定各因素的代表性指标; 二是运用机器学习方法, 选取CatBoost算法对其进行贝叶斯优化, 据此预测铁矿石资源供应系数的未来走势, 同时与其他机器学习方法的预测结果进行对比; 三是采用SHAP方法解释模型结果, 对比分析各个因素系数对于铁矿石产业链供应链安全性的贡献水平。研究结论如下: BO-Catboost模型预测供应系数的效果要优于其他机器学习方法。在铁矿石供应系数的影响因素特征排名中, 经济风险相关指标权重最大, 其次是经济发展因素、运输周转因素、价格波动因素的指标权重最小。其中, 经济不确定性指数、黑色金属冶炼和压延加工业亏损企业数量比例、美元兑人民币汇率与铁矿石库存量是影响2012—2024年间铁矿石资源供应的主要原因。中国铁矿石资源供应面临诸多风险, 需要根据国际市场供需情况与经济发展水平, 灵活调整铁矿石进口关税和进口节奏, 保持市场稳定。基于相关风险防范, 本文为政府和企业提供了优化铁矿石资源配置和应对经济环境风险的政策建议。 |
| 中文关键词:铁矿石资源供应 特征分析 SHAP模型 |
| |
| Evaluation of China’s Iron Ore Resource Industry Chain Based on Machine Learning |
|
|
| Abstract:Due to the low grade and insufficient supply of domestic iron ore resources, China’s iron smelting production is highly dependent on imported iron ore. Therefore, it is particularly important to scientifically assess the security situation of China’s iron ore resources. This study quantitatively evaluates the supply security of China’s iron ore resources, provides real-time early warnings, and proposes corresponding strategies and measures for existing risk issues. First, relevant indicators are selected and constructed from the aspects of iron ore security guarantee, eco-nomic development, transportation turnover, economic risk, and price fluctuation factors. Second, the CatBoost algorithm is selected, optimized through Bayesian optimization using machine learning methods to predict the supply security coefficient of iron ore resources, and compared with other machine learning methods. Finally, the SHAP method is used to explain the model results and compare and analyze the contribution levels of each factor coefficient to the security of the iron ore industry and supply chains. Results show that the BO–Catboost model is more effective in predicting the supply coefficient than other machine learning methods. In the ranking of the characteristics of the factors influencing the iron ore coefficient, the economic risk-related indicators have the largest weight, followed by those of the economic development and transportation turnover factors, while the price fluctu-ation factors have the smallest indicator weight. Between 2012 and 2024, the economic uncertainty index, proportion of loss-making enterprises in the ferrous metal smelting and rolling processing industry, USD–CNY exchange rate, and iron ore inventory were the main aspects that affected the supply security of iron ore resources. China’s iron ore resource supply faces many risks and challenges, among which the economic environment risk has the most sig-nificant impact. To maintain market stability and supply security, it is necessary to flexibly adjust the import tariffs and import rhythm of iron ore according to the international market supply and demand situation and the level of economic development. Based on relevant risk indicators, this study provides policy suggestions for the government and enterprises to optimize the allocation of iron ore resources and deal with economic environment risks. |
| keywords:iron ore resources supply feature analysis SHAP |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
|
|
