多层时序网络视角下的最优投资组合策略研究

doi:10.16381/j.cnki.issn1003-207x.2023.0851

摘要/Abstract

摘要：

从股票资产间的线性和非线性关联性及动态演化特征出发理解股票市场，对于投资组合优化研究具有重要的意义。本文将多层时序网络与最优化理论相结合，以多层时序网络特征向量中心性测度为基础，设计网络风险度量指标，创新性地提出全局最小网络风险投资组合模型，基于2010—2022年沪深300成分股数据，模拟动态投资过程并结合多种评价指标评估投资组合模型。研究结果表明：多层时序网络可综合探明股票资产的关联性结构和演变特征，准确刻画复杂金融系统的结构，识别出优质的投资资产；边缘投资组合模型可以获得更好的投资绩效及累积收益率，且这种收益不被系统性风险因子暴露所抵消；边缘全局最小网络风险投资组合模型在外样本期间有着最优的投资表现，且在股市波动时依然保持较强的稳健性，适合风险承受能力弱的投资者使用。研究结论丰富了投资者的投资策略，尤其是对风险承受能力较弱的散户有一定的参考意义。

关键词: 多层时序网络, 股票市场, 投资组合优化

Abstract:

Stocks of varying quality have put forward new requirements for investors’ investment research capabilities， especially for Chinese investors who are mainly natural persons （retail investors）. Since retail investors have weak risk tolerance and high degree of loss aversion， facing the increasing asset scale， how to obtain higher returns with the lowest possible risk is undoubtedly the issue that investors are most concerned about. The stock market is studied from the linear and nonlinear correlation and dynamic evolution characteristics between stock assets. First， it describes the stock market with multilayer temporal network， and designs network risk measurement indicators based on the eigenvectors centrality measure. Then Combining it with optimization theory， the Global Minimum Network-Risk portfolio model is proposed to simulate dynamic investment process based on the data of HS300 constituent stocks from 2010 to 2022， and then the portfolio model with a variety of evaluation indicators is evaluated.Experimental results are as follows （i）The multilayer temporal network can comprehensively ascertain the correlation structure and evolution characteristics of stock assets， accurately describe the structure of complex financial systems， and identify high-quality investment assets；（ii）The peripheral portfolio model can obtain better investment performance and cumulative return rate， and the return is not offset by systematic risk factor exposure；（iii）The peripheral Global Minimum Network-Risk portfolio model has the best investment performance during the out-of-sample period， and it still maintains strong robustness in stock market fluctuations， which is suitable for investors with low risk tolerance. Theoretical and practical references are provided for investors with different risk preferences especially for retail investors with low risk tolerance， the traditional mean-variance portfolio theory is expanded and improved based on multilayer temporal network， and further research in this field is enriched.

Key words: multilayer temporal network, stock market, portfolio optimization

中图分类号:

F832.5

刘超, 许澜涛. 多层时序网络视角下的最优投资组合策略研究[J]. 中国管理科学, 2025, 33(9): 46-56.

Chao Liu, Lantao Xu. Study on Optimal Portfolio Strategy from the Perspective of Multilayer Temporal Network[J]. Chinese Journal of Management Science, 2025, 33(9): 46-56.

图/表 11

图1

表1

GMCN中η取值及投资者的态度"

$η$	投资者态度
$η = 0.25$	较厌恶CVaR
$η = 0.5$	同等厌恶CVaR和网络风险
$η = 0.75$	较厌恶网络风险
$η = 1$	只厌恶网络风险

表1

图2

图3

图4

图5

表2

投资组合在一个持有期的投资绩效"

股票数	投资组合		SR	OR	Beta	Alpha
n=5	HS300		0.179	1.534	1.000
	(1/n)_C		-0.765	0.720	1.438	-0.186
	(1/n)_P		0.437	3.239	3.024	0.354
	GMC_C		0.027	1.215	0.874	-0.025
	GMC_P		-0.036	0.843	2.063	-0.166
	GMCN_C	$η = 0.25$	-0.145	0.644	1.304	-0.235
		$η = 0.5$	-0.083	0.807	1.394	-0.148
		$η = 0.75$	-0.048	0.888	1.422	-0.115
	GMCN_P	$η = 0.25$	0.274	2.235	3.195	0.173
		$η = 0.5$	0.404	2.957	3.079	0.319
		$η = 0.75$	0.430	3.350	2.998	0.385
	GMN_C	$η = 1$	-0.026	0.939	1.438	-0.097
	GMN_P	$η = 1$	0.444	3.600	2.952	0.424
n=10	(1/n)_C		-0.100	0.710	1.510	-0.175
	(1/n)_P		0.268	1.982	3.018	0.180
	GMC_C		-0.198	0.250	0.483	-0.286
	GMC_P		0.246	2.245	1.875	0.226
	GMCN_C	$η = 0.25$	0.195	1.931	2.209	0.112
		$η = 0.5$	0.163	1.555	2.217	0.033
		$η = 0.75$	0.125	1.364	2.164	-0.012
	GMCN_P	$η = 0.25$	0.274	2.235	3.175	0.173
		$η = 0.5$	0.337	2.702	4.030	0.267
		$η = 0.75$	0.242	1.862	3.728	0.100
	GMN_C	$η = 1$	0.094	1.270	2.127	-0.035
	GMN_P	$η = 1$	0.157	1.518	3.553	0.008
n=15	(1/n)_C		-0.028	0.993	1.653	-0.096
	(1/n)_P		0.310	2.238	3.217	0.175
	GMC_C		0.136	1.498	1.235	0.050
	GMC_P		0.448	3.524	1.764	0.433
	GMCN_C	$η = 0.25$	0.080	1.220	1.810	0.013
		$η = 0.5$	0.165	1.502	1.804	0.061
		$η = 0.75$	0.193	1.619	1.912	0.059
	GMCN_P	$η = 0.25$	0.491	4.555	2.525	0.445
		$η = 0.5$	0.360	2.664	3.104	0.232
		$η = 0.75$	0.274	2.053	3.544	0.131
	GMN_C	$η = 1$	0.164	1.543	2.017	0.037
	GMN_P	$η = 1$	0.226	1.797	3.800	0.081

表2

图6

表3

外样本期间Portfolio_P的投资绩效"

模型		SR	OR	Beta	Alpha
HS300		-0.004	0.990	1.000
(1/n)_P		0.044	1.128	1.193	0.064
GMC_P		-0.004	0.979	1.175	-0.008
GMCN_P	$η = 0.25$	0.018	1.056	1.213	0.031
	$η = 0.5$	0.035	1.102	1.194	0.052
	$η = 0.75$	0.043	1.125	1.178	0.062
GMN_P	$η = 1$	0.046	1.141	1.167	0.069

表3

图7

图8

参考文献 37

[1]	Markowitz H. Portfolio selection［J］. The Journal of Finance， 1952， 7（1）： 77-91.
[2]	Li P， Han Y， Xia Y. Portfolio optimization using asymmetry robust mean absolute deviation model［J］. Finance Research Letters， 2016， 18： 353-362.
[3]	Kang Z， Li X， Li Z， et al. Data-driven robust mean-CVaR portfolio selection under distribution ambiguity［J］. Quantitative Finance， 2019， 19（1）： 105-121.
[4]	徐维军，周平平，李婷，等. 基于CVaR和多元权值约束下的积极投资组合模型［J］. 系统管理学报， 2017， 26（2）： 219-224+233.
	Xu W J， Zhou P P， Li T， et al. Active portfolio model with CVaR and multiple weight constraints［J］. Journal of Systems & Management， 2017， 26（2）： 219-224+233.
[5]	Gao J， Xiong Y， Li D. Dynamic mean-risk portfolio selection with multiple risk measures in continuous-time［J］. European Journal of Operational Research， 2016， 249（2）： 647-656.
[6]	Roman D， Darby-Dowman K， Mitra G. Mean-risk models using two risk measures： A multi-objective approach［J］.Quantitative Finance，2007，7（4）： 443-458.
[7]	Zhao X， Shi Y. Asset allocation and reinsurance policy for a mean-variance-CVaR insurer in continuous-time［J］. Chinese Journal of Applied Probability and Atatistics， 2020，36（5）： 536-550
[8]	欧阳红兵，刘晓东. 中国金融机构的系统重要性及系统性风险传染机制分析——基于复杂网络的视角［J］. 中国管理科学， 2015， 23（10）： 30-37.
	Ouyang H B， Liu X D. An analysis of the systemic importance and systemic risk contagion mechanism of China’s financial institutions based on network analysis［J］. Chinese Journal of Management Science， 2015， 23（10）： 30-37.
[9]	刘超，高凤凤，张梦婉，等. 中国金融市场风险溢出效应、冲击效应与风险预警研究［J］. 中国管理科学， 2025， 33（5）： 1-12.
	Liu C， Gao F F， Zhang M W， et al. Research on risk spillover effect， impact effect and risk early warning in China's financial market［J］. Chinese Journal of Management Science， 2025， 33（5）： 1-12.
[10]	刘超，钱存. 多目标条件下金融监管系统优化与风险管理［M］. 北京：中国财政经济出版社， 2021.
	Liu C， Qian C. Optimization of financial supervision system and risk management under multi-objective conditions［M］. Beijing： China Financial and Economic Publishing House， 2021.
[11]	刘超，徐君慧，周文文. 中国金融市场的风险溢出效应研究——基于溢出指数和复杂网络方法［J］. 系统工程理论与实践， 2017， 37（4）： 831-842.
	Liu C， Xu J H， Zhou W W. Study on risk spillover effect of financial markets in China based on methods of spillover index and complex network［J］.Systems Engineering-Theory & Practice，2017，37（4）： 831-842.
[12]	Mantegna R N. Hierarchical structure in financial markets［J］. The European Physical Journal B - Condensed Matter and Complex Systems， 1999， 11（1）： 193-197.
[13]	Pozzi F， Di Matteo T， Aste T. Spread of risk across financial markets： Better to invest in the peripheries［J］. Scientific Reports， 2013（3）： 1-7.
[14]	Peralta G， Zareei A. A network approach to portfolio selection［J］. Journal of Empirical Finance， 2016， 38： 157-180.
[15]	Li Y， Jiang X F， Tian Y， et al. Portfolio optimization based on network topology［J］. Physica A： Statistical Mechanics and Its Applications， 2019， 515： 671-681.
[16]	Yang L， Zhao L， Wang C. Portfolio optimization based on empirical mode decomposition［J］. Physica A： Statistical Mechanics and Its Applications， 2019， 531： 121813.
[17]	Zhao L， Wang G J， Wang M， et al. Stock market as temporal network［J］. Physica A： Statistical Mechanics and Its Applications， 2018， 506： 1104-1112.
[18]	赵霞，许澜涛，孙晓，等. 带耦合时序网络、重要性节点识别及投资组合研究—以股票市场为例［J］. 应用概率统计， 2023， 39（1）： 117-131.
	Zhao X， Xu L T， Sun X， et al. Study on temporal network with coupling， nodes importance and portfolio optimization： A case of stock market［J］. Chinese Journal of Applied Probability and Statistics， 2023， 39（1）： 117-131.
[19]	Battiston F， Nicosia V， Latora V. Structural measures for multiplex networks［J］. Physical Review E， Statistical， Nonlinear， and Soft Matter Physics， 2014， 89（3）： 032804.
[20]	Wang G J， Wan L， Feng Y， et al. Interconnected multilayer networks： Quantifying connectedness among global stock and foreign exchange markets［J］. International Review of Financial Analysis， 2023， 86： 102518.
[21]	张飞鹏，徐一雄，邹胜轩，等. 基于LGCNET多层网络的中国A股上市公司系统性风险度量［J］. 中国管理科学， 2022， 30（12）： 13-25.
	Zhang F P， Xu Y X， Zou S X， et al. An empirical study on the systemic risk of Chinese A-share listed companies based on multi-layer network［J］. Chinese Journal of Management Science， 2022， 30（12）： 13-25.
[22]	Gong J， Wang G J， Zhou Y， et al. Spreading of cross-market volatility information： Evidence from multiplex network analysis of volatility spillovers［J］. Journal of International Financial Markets， Institutions and Money， 2023， 83： 101733.
[23]	李守伟，王虎，刘晓星. 基于银行动态多层网络的系统性风险防控政策效果研究［J］. 管理工程学报， 2022， 36（4）： 164-176.
	Li S W， Wang H， Liu X X. Research on the effect of systemic risk prevention and control policy based on dynamic multilayer network of banks［J］. Journal of Industrial Engineering and Engineering Management， 2022， 36（4）： 164-176.
[24]	王纲金，吴昊钰，谢赤. 基于多层关联网络的投资组合优化研究［J］. 系统工程理论与实践， 2022， 42（4）： 937-957.
	Wang G J， Wu H Y， Xie C. Portfolio optimization based on multilayer connectedness networks［J］. Systems Engineering-Theory & Practice， 2022， 42（4）： 937-957.
[25]	Arrigo F， Tudisco F. Multi-dimensional， multilayer， nonlinear and dynamic HITS［C］//Proceedings of the 2019 SIAM International Conference on Data Mining， Calgary，AB，CAN，May 2-4，SIAM，2019： 369-377.
[26]	Wang D， Yu W， Zou X. Tensor-based mathematical framework and new centralities for temporal multilayer networks［J］. Information Sciences， 2020， 512： 563-580.
[27]	Zhang T， Zhang K， Lv L， et al. Co-Ranking for nodes， layers and timestamps in multilayer temporal networks［J］. Chaos， Solitons & Fractals， 2019， 125： 88-96.
[28]	Lv L， Zhang K， Bardou D， et al. HITS centrality based on inter-layer similarity for multilayer temporal networks［J］. Neurocomputing， 2021， 423： 220-235.
[29]	Massara G P， Di Matteo T， Aste T. Network filtering for big data： Triangulated maximally filtered graph［J］. Journal of Complex Networks， 2017， 5（2）： 161-178.
[30]	Merton R C. On estimating the expected return on the market An exploratory investigation［J］. Journal of Financial Economics， 1980， 8（4）： 323-361.
[31]	汪寿阳，陈实，乔晗，等. 公司风险投资的分散与过度投资行为——基于行业间相互投资网络的绩效评估［J］. 南开管理评论， 2021， 24（5）： 128-140.
	Wang S Y， Chen S， Qiao H， et al. The decentralization and overinvestment behavior of CVC performance evaluation based on the investment network relationship between industries［J］. Nankai Business Review， 2021， 24（5）： 128-140.
[32]	杨子晖，陈雨恬，林师涵. 系统性金融风险文献综述：现状、发展与展望［J］. 金融研究， 2022（1）： 185-206.
	Yang Z H， Chen Y T， Lin S H. A literature review of systemic risk： Status， development and prospect［J］. Journal of Financial Research， 2022（1）： 185-206.
[33]	Clemente G P， Grassi R， Hitaj A. Asset allocation： New evidence through network approaches［J］. Annals of Operations Research， 2021， 299（1）： 61-80.
[34]	莫东序，郑田丹. 基于复杂网络的投资组合优化研究［J］. 中国管理科学， 2021， 29（5）： 25-33.
	Mo D X， Zheng T D. Research on portfolio optimization based on complex network［J］. Chinese Journal of Management Science， 2021， 29（5）： 25-33.
[35]	DeMiguel V， Garlappi L， Uppal R. Optimal versus naive diversification： How inefficient is the 1/N portfolio strategy？［J］. The Review of Financial Studies， 2009， 22（5）： 1915-1953.
[36]	Sharpe W F. Capital asset prices： A theory of market equilibrium under conditions of risk［J］. The Journal of Finance， 1964， 19（3）： 425-442.
[37]	徐凤敏，刘文玲，李雪鹏，等. “双碳” 目标下绿色双层投资组合研究［J］.统计与信息论坛，2023，38（1）： 55-70.
	Xu F M， Liu W L， Li X P， et al. Research on green bilevel portfolio under carbon peaking and carbon neutrality goals［J］. Journal of Statistics and Information， 2023， 38（1）： 55-70.