beoplay体育提现学报(社会科学版) ›› 2019, Vol. 29 ›› Issue (3): 57-65.

• 经济学 • 上一篇    下一篇

基于混合结构数据的碳价格多尺度组合预测

任贺松1,2,刘金培1,3,郭艺1,4,郭健1,5   

  1. 1. 安徽大学商学院,安徽合肥230601
    2. 西安交通大学管理学院,陕西西安710049
    3. 北卡罗莱纳州立大学工业与系统工程系,美国罗利 27695
    4. 东南大学经济管理学院,江苏南京2111895
  • 收稿日期:2019-07-23 出版日期:2019-09-25 发布日期:2019-09-25
  • 作者简介:任贺松(1994-),男,安徽界首人,西安交通大学,博士,研究方向:预测与决策分析|刘金培(1984-),男,山东滨州人,安徽大学商学院,副教授,研究方向:预测与决策分析
  • 基金资助:
    国家自然科学基金(71901001);国家自然科学基金(71501002);国家自然科学基金(71871001);安徽省自然科学基金杰出青年基金(1908085J03);安徽省高校人文社科基金重点项目(SK2019A0013);安徽省高校人文社科基金重点项目(SK2018A0605)

Multiscale Combined Forecast of Carbon PriceBased on Mixed Structure Data

REN HESONG1,2,LIU JINPEI1,3,GUO YI1,4,GUO JIAN1,5   

  1. 1. School of Business, Anhui University, Hefei, Anhui, 230601, China
    2. School of Management, Xi'an Jiaotong University, Xi’an, Shanxi, China;
    3. Edward P. Fitts Department of Industrial and Systems Engineering, North Carolina State University, Raleigh, NC, 27695, USA
    4. School of Economics and Management, Southeast University, Nanjing, Jiangsu, 211189, China5. School of Economics and Management, Tongji University, Shanghai 200092,China
  • Received:2019-07-23 Online:2019-09-25 Published:2019-09-25

摘要:

碳交易价格的精准预测对推动碳交易市场的科学理性发展具有重要意义,因此提出一种基于混合结构数据的碳价格多尺度组合预测方法。首先,用谷歌指数提取碳价格相关的非结构化数据,基于主成分分析对其进行降维。然后,对影响因素的结构化数据、降维后的非结构化数据、碳交易价格分别进EMD分解得到不同个数的本征模函数(IMF),并采用Fine-to-Coarse技术对IMF进行重构得到高频序列、低频序列和趋势项。进而根据时间序列各尺度特点,用ARIMA、PLS 和神经网络对高频数据、低频数据和趋势项进行预测。最后,对预测结果集成得到碳价格预测序列。以欧盟碳价格为例进行实证分析,结果表明,此组合预测模型的预测精度优于单项预测方法和未对时间序列进行EMD分解处理的预测方法,具有较好适用性。

关键词: EMD分解, 组合预测, 碳价格, PLS, 非结构化数据

Abstract:

Accurate forecast of the carbon trading price is of great significance in promoting the scientific and rational development of carbon trading market. Therefore, this paper proposes a multi-scale combined forecasting method for carbon price based on mixed structure data. First, the Google Index is used to extract the unstructured data related to the carbon price.The dimensions of unstructured data are reduced based on principal component analysis. Then, EMD is employedto the structured data,unstructured data and the carbon trading price to obtain different IMFs, which are reconstructed by the Fine-to-Coarse technique to get low, high frequency sequence and trend sequence. Furthermore, the three items are predicted respectively by using ARIMA, PLS and neural networks according to the features of each scale in time series. Finally, the forecasting results are summed to get the carbon price forecast sequence. The proposed method is used to forecast carbon price in EU. The empirical results show that the prediction accuracy of the model is higher than that of the single prediction method and the prediction method that time series aren’t decomposed by EMD, which is of great applicability.

Key words: EMD decomposition, combination forecast, carbon price, PLS, unstructured data

中图分类号: 

  • F407.21