主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2024, Vol. 32 ›› Issue (4): 1-13.doi: 10.16381/j.cnki.issn1003-207x.2021.0382

• •    

一个基于BL模型和复杂网络的行业配置模型

李仲飞1,周骐2()   

  1. 1.南方科技大学金融系,广东 深圳 518055
    2.华南理工大学工商管理学院,广东 广州 510641
  • 收稿日期:2021-02-26 修回日期:2021-04-20 出版日期:2024-04-25 发布日期:2024-05-11
  • 通讯作者: 周骐 E-mail:zhouqi@scut.edu.cn
  • 基金资助:
    国家自然科学基金项目(71991474);广东省哲学社会科学规划项目(GD23XGL093);广东省自然科学基金项目(2023A1515010597);广州市基础与应用基础项目(2024A04J3762)

An Industry Allocation Model Based on BL Model and Complex Network

Zhongfei Li1,Qi Zhou2()   

  1. 1.Department of Finance, Business School, Southern University of Science and Technology, Shenzhen 518055, China
    2.School of Business Administration, South China University of Technology, Guangzhou 510641, China
  • Received:2021-02-26 Revised:2021-04-20 Online:2024-04-25 Published:2024-05-11
  • Contact: Qi Zhou E-mail:zhouqi@scut.edu.cn

摘要:

行业配置是投资组合中承上启下的重要环节之一。传统的行业配置模型忽略了行业网络的整体关联性。本文以行业配置中广泛应用的Black-Litterman(BL)模型为基础(以机器学习和时间序列模型的预测值为主观观点),结合复杂网络方法,提出了BL模型的图示化表达方式,证明了BL模型最优投资组合权重与行业网络特征向量中心度之间的二次关系并进行了实证分析。在此基础上,提出了BL+Network行业配置模型,综合考虑行业间的整体关联风险,然后确定各行业投资比例。该过程从行业配置层面完善了复杂网络方法在“自上而下”投资组合管理中的应用框架。从与传统资产配置模型的实证比较发现,BL+Network模型的行业配置在夏普比率和增益损失比等指标上均有明显提高,在特征向量中心度处于序列中位数附近的行业配比较高,但配置行业数量并非越少越好。本研究为资产配置的研究提供了新视角,拓展了BL模型的交叉应用边界,补充了复杂网络方法在投资组合管理中的应用。

关键词: 复杂网络, BL+Network, 特征向量中心度

Abstract:

Industrial allocation is one of the indispensable steps in portfolio investment process. The traditional allocation models of industry ignores the overall correlation of the industry network. Based on the Black-Litterman (BL) model (the predictive values of machine learning and time series models are subjective points of view) in industrial allocation together with complex network theory, a graphical presentation of the BL model is proposed, the mathematical relation between the optimal portfolio of the BL model and the graphical presentation is proved and the corresponding empirical analysis is done. Then an industry allocation model is proposed, named BL+Network model. The model perfects the application framework of complex network method in “top-down” portfolio management. Empirical comparison with some traditional models finds that BL+Network model improves evidently Sharpe ratio and gain-loss ratio, industries whose eigenvector centrality degree are in near of the median sequence number should be allocated more, but fewer industries is not always better. This research provides a new perspective for asset allocation study, and extends the cross application of BL model and complex network method to portfolio management.

Key words: complex network, BL+Network model, eigenvector centrality degree

中图分类号: