基于Black-Litterman模型的行业资产配置投资方案研究

随着我国经济实力的发展，我国国民可支配收入持续增加，居民对于资产配置的需求日益凸显。根据Wind公开数据，近十年基金市场整体资产规模增长近12倍，年几何平均增长率高达27.93%。而我国基金市场的资产配置理论还处在初级发展阶段，相比于西方发达市场，资产量化分析的市场占比还非常低，缺乏行之有效的行业资产配置策略。

而随着机器学习理论的发展和硬件的升级，神经网络在金融时间序列数据处理上的应用更是成为了近些年来的研究热点。2014年提出的门控循环单元（GRU）神经网络模型，能够通过门控逻辑和循环结构处理非线性、非平稳的时间序列，并记忆序列的长期信息，解决梯度消失的问题。因此，将新型神经网络模型和资产配置模型结合指导投资组合的构建也就成为了有探索价值的研究方向。

本文在传统的Black-Litterman资产配置模型的基础上，开创性地引入了GRU神经网络算法预测资产收益率从而生成投资者观点向量，指导行业资产配置。在实证上，本文选择了全面覆盖中国A股市场的中证全指一级行业指数作为标的资产，通过GRU模型预测行业指数的预期收益率，并结合模型训练效果使用模型平均绝对误差生成投资者观点信心水平矩阵，进而结合行业指数在市场均衡状态下的先验收益分布，通过BL模型中的贝叶斯收缩推断出后验预期收益率的分布，最后通过二次规划逆推出最优资产配置比例，构建了行业指数投资组合。

本文开创性构造的GRU-BL模型，解决了传统BL模型投资者主观观点选择过于随意且参数设置过于简单以及时间序列类BL模型对非线性不平稳收益序列拟合效果不佳且存在梯度消失现象的问题。最终的实证结果表明，GRU-BL模型给出的最优资产配置权重在真实市场上的超额收益率及夏普比率均高于均值方差模型和市场权重模型，整体表现优异。

With the rapid development of China's economy and the continuous increase of national disposable income, residents' demand for asset allocation has become increasingly prominent. According to Wind's public data, the overall asset size of the Chinese mainland fund market has increased by nearly 12 times in the past decade, with an annual geometric average growth rate of 27.93%. However, the asset allocation theory of our country's fund market is still in the initial stage of development, and the market share of quantitative analysis is still small compared to developed western markets, and there is a lack of effective industry asset allocation strategies. 

With the development of machine learning theory and the upgrade of hardware, the application of neural network in financial time series data processing has become a research hotspot in recent years. The Gated Recurrent Unit (GRU) neural network model, which proposed in 2014, can process non-linear and non-stationary time series data through gated logic and cyclic structure, and memorize long-term information of the sequence to solve the vanishing gradient problem. Therefore, the combination of the new neural network model and the asset allocation model to guide the construction of the investment portfolio has become a research direction with exploratory value.

Based on the traditional Black-Litterman asset allocation model, this paper innovatively introduces the GRU neural network algorithm to predict asset returns to generate investor view vectors, which ultimately guide industry asset allocation. Empirically, this paper selects the CSI Tier 1 Industry Indexes, which comprehensively covers China's A-share market, as the underlying asset. This paper uses the GRU model to predict the expected return of the industry indexes and combines the model training effect to generate the investor view confidence level matrix using the model mean absolute error. Subsequently, combined with the prior return distribution of the industry indexes in the market equilibrium state, the distribution of the posterior expected return is inferred through the Bayesian contraction in the BL model. Finally, the optimal asset allocation ratio is inversely derived through quadratic linear programming, and an industry index portfolio is constructed.

The groundbreaking GRU-BL model constructed in this paper solves the problem that the traditional BL model investors choose too arbitrarily, the parameter setting is too simple, and the time series BL model has poor fitting effect on nonlinear or non-stationary income series coupled with vanishing gradient phenomenon. The final empirical results show that the excess return and Sharpe ratio of the optimal asset allocation weight given by the GRU-BL model in the real market are higher than those of the mean variance model and the market weight model.

[1] MARKOWITZ H M. Portfolio Selection, 1959[J]. Journal of Finance, 1952, 7: 7791.
[2] CHOPRA V K, HENSEL C R, TURNER A L. Massaging mean-variance inputs: returns from alternative global investment strategies in the 1980s[J]. Management Science, 1993, 39(7): 845-855.
[3] CHOUEIFATY Y, COIGNARD Y. Toward maximum diversification[J]. The Journal of Port- folio Management, 2008, 35(1): 40-51.
[4] MICHAUD R O. The Markowitz optimization enigma: Is‘optimized’optimal?[J]. Financial analysts journal, 1989, 45(1): 31-42.
[5] 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.
[6] FAMA E F, FRENCH K R. The cross-section of expected stock returns[J]. the Journal of Finance, 1992, 47(2): 427-465.
[7] KING B F. Market and industry factors in stock price behavior[J]. the Journal of Business, 1966, 39(1): 139-190.
[8] LIVINGSTON M. Industry movements of common stocks[J]. The Journal of Finance, 1977, 32(3): 861-874.
[9] MOSKOWITZ T J, GRINBLATT M. Do industries explain momentum?[J]. The Journal of finance, 1999, 54(4): 1249-1290.
[10] ISAKOV D, SONNEY F. Are practitioners right? On the relative importance of industrial factors in international stock returns[J]. On the Relative Importance of Industrial Factors in International Stock Returns (February 2003), 2003.
[11] GREETHAM T, HARTNETT H. The Investment Clock Special Report# 1: Making Money From Macro[J]. Merrill Lynch, 2004.
[12] 蒋治平. 我国股市行业指数之间的冲击传导研究[J]. 證券市場導報, 2008, 2008(10): 23-28.
[13] 陈暮紫, 陈敏, 吴武清, 等. 中国 A 股市场行业板块间领滞关系的动态变化实证研究[J]. 系统工程理论与实践, 2009, 29(6): 19-31.
[14] 陈波. 基于货币周期的行业轮动策略在我国股市的实证分析[J]. 企业导报, 2011(11): 137-137.
[15] 张羽乔. 我国股票市场的行业轮动性分析[J]. 中国商论, 2017(32): 34-36.
[16] BLACK F, LITTERMAN R. Global portfolio optimization[J]. Financial analysts journal, 1992, 48(5): 28-43.
[17] BLACK F, LITTERMAN R. Asset allocation: combining investor views with market equilib- rium[J]. Goldman Sachs Fixed Income Research, 1990, 115.
[18] LITTERMAN R, HE G. The intuition behind black-litterman model portfolios[J]. Goldman Sachs Investment Management Research, 1999.
[19] BEVAN A, WINKELMANN K. Using the Black-Litterman global asset allocation model: three years of practical experience[J]. Fixed Income Research, 1998: 1-19.
[20] LEE W. Theory and methodology of tactical asset allocation: volume 65[M]. John Wiley & Sons, 2000.
[21] SATCHELL S, SCOWCROFT A. A demystification of the Black-Litterman model: Manag- ing quantitative and traditional portfolio construction[M]//Forecasting Expected Returns in the Financial Markets. Elsevier, 2007: 39-53.
[22] BEACH S L, ORLOV A G. An application of the Black–Litterman model with EGARCH- M-derived views for international portfolio management[J]. Financial Markets and Portfolio Management, 2007, 21(2): 147-166.
[23] LEJEUNE M A. A VaR Black–Litterman model for the construction of absolute return fund- of-funds[J]. Quantitative Finance, 2011, 11(10): 1489-1501.
[24] MEUCCI A. Beyond Black-Litterman in practice: A five-step recipe to input views on non- normal markets[J]. Available at SSRN 872577, 2006.
[25] MEUCCI A. Enhancing the Black–Litterman and related approaches: Views and stress-test on risk factors[J]. Journal of Asset Management, 2009, 10(2): 89-96.
[26] XIAO Y, VALDEZ E A. A Black–Litterman asset allocation model under Elliptical distributions [J]. Quantitative Finance, 2015, 15(3): 509-519.
[27] 郭红. 运用 Black-Litterman 模型对我国股票市场行业配置研究[D]. 首都经济贸易大学, 2012.
[28] 孟勇. 对 Black-Litterman 模型加入主观收益方法的改进[J]. 统计研究, 2012(2): 94-99.
[29] 汤志坚. 基于改进 Black-Litterman 模型的证券资产配置研究[D]. 大连理工大学, 2013.
[30] 柯海平. 基于 Black-Litterman 模型的行业资产配置研究[D]. 哈尔滨工业大学, 2017.
[31] 李宗铭, 房勇. 基于 LSTM 神经网络的行业资产配置模型[J]. 系统工程理论与实践, 2021: 1.
[32] 陈浩南. 基于 LSTM 神经网络算法的 Black-Litterman 投资组合优化及应用研究[D]. 成都 理工大学, 2020.
[33] HOPFIELD J J. Neural networks and physical systems with emergent collective computational abilities[J]. Proceedings of the national academy of sciences, 1982, 79(8): 2554-2558.
[34] ELMAN J L. Finding structure in time[J]. Cognitive science, 1990, 14(2): 179-211.
[35] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural computation, 1997, 9(8): 1735-1780.
[36] CHO K, VAN MERRIËNBOER B, GULCEHRE C, et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation[J]. arXiv preprint arXiv:1406.1078, 2014.
[37] CHUNG J, GULCEHRE C, CHO K, et al. Empirical evaluation of gated recurrent neural net- works on sequence modeling[J]. arXiv preprint arXiv:1412.3555, 2014.
[38] WHITE H. Economic prediction using neural networks: The case of IBM daily stock returns [C]//ICNN: volume 2. 1988: 451-458.
[39]  HINTON G E, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J]. Neural computation, 2006, 18(7): 1527-1554.
[40]  禹建丽, 孙增圻, 刘治军, 等. 基于 BP 神经网络的股市建模与决策[J]. 系统工程理论与实 践, 2003, 23(5): 15-19.
[41]  于志军, 杨善林, 章政, 等. 基于误差校正的灰色神经网络股票收益率预测[J]. 中国管理科 学, 2015, 23(12): 20-26.
[42]  牛红丽, 赵亚枝. 利用 Bagging 算法和 GRU 模型预测股票价格指数[J]. 计算机工程与应 用, 2021: 1-9.