Skip to main content
SpringerLink
Log in

Portfolio construction using bootstrapping neural networks: evidence from global stock market

  • Published:
Review of Derivatives Research Aims and scope Submit manuscript

Abstract

The study investigates the investment value of global stock markets by a portfolio construction method combined with bootstrapping neural network architecture. A residual sample will be generated from bootstrapping sample procedure and then incorporated into the estimation of the expected returns and the covariant matrix. The outputs are further processed by the traditional Markowitz optimization procedure. In order to examine the efficacy of the proposed approach, the illustrated case was compared with traditional Markowitz mean–variance analysis, as well as the James–Stein and minimum-variance estimators. From the empirical results, it indicated that this novel approach significantly outperforms most of benchmark models based on various risk-adjusted performance measures. It can be shown that this new approach has great promise for enhancing the estimation of the investment value by Markowitz mean–variance analysis in the global stock markets.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2

Similar content being viewed by others

References

  • Board, J., & Sutcliffe, C. (1994). Estimation methods in portfolio selection and the effectiveness of short sales restrictions: UK evidence. Management Science, 40(4), 516–534.

    Article  Google Scholar 

  • Chang, E., Eun, C. S., & Kolodny, R. (1995). International diversification through closed-end country funds. Journal of Banking & Finance, 17, 1237–1263.

    Article  Google Scholar 

  • Davidson, R., & MacKinnon, T. (2004). Econometric theory and methods. New York: Oxford University Press.

    Google Scholar 

  • Efron, B., & Tibshirani, R. (1993). An introduction to the bootstrap. New York: Chapman and Hall.

    Book  Google Scholar 

  • Eiteman, D. K., Stonehill, A. T., & Moffett, M. H. (1998). Multinational business finance (8th ed.). Boston: Addison Wesley.

    Google Scholar 

  • Elman, J. L. (1990). Finding structure in time. Cognitive Science, 14, 179–211.

    Article  Google Scholar 

  • Fernandez, A., & Gomez, S. (2007). Portfolio selection using neural networks. Computers & Operations Research, 34, 1177–1191.

    Article  Google Scholar 

  • Franke, J., & Klein, M. (1999). Optimal portfolio management using neural networks: A case study. Kaiserslautern, Germany: University of Kaiserslautern, Department of Mathematics Technical Report.

    Google Scholar 

  • Franke, J., & Neumann, M. (2000). Bootstrapping neural networks. Neural Computation, 12(8), 1929–1949.

    Article  Google Scholar 

  • Freedman, D. A. (1982). Bootstrapping regression models. Annual Statistics, 9, 1218–1228.

    Article  Google Scholar 

  • Frost, P. A., & Savarino, J. E. (1986). An empirical Bayes approach to efficient portfolio selection. Journal of Financial and Quantitative Analysis, 21(3), 293–305.

    Article  Google Scholar 

  • Giles, C. L., Lawrence, A., & Tsoi, A. C. (2001). Noisy time series prediction using a recurrent neural network and grammatical inference. Machine Learning, 44, 161–183.

    Article  Google Scholar 

  • Graham, J. R., & Harvey, C. R. (1997). Grading the performance of market-timing newsletters. Financial Analysts Journal, 53(6), 54–66.

    Article  Google Scholar 

  • Grubel, H. G. (1968). Internationally diversified portfolios: Welfare gains and capital flows. American Economic Review, 58, 1299–1314.

    Google Scholar 

  • Hall, P. (1992). The bootstrap and edgeworth expansion. Berlin: Springer.

    Book  Google Scholar 

  • Haykin, S. (1999). Neural networks: A comprehensive foundation. Upper Saddle River: Prentice-Hall Inc.

    Google Scholar 

  • Huang, C. S., Lin, Z. W., & Chen, C. W. (2012). Bootstrapping multilayer neural networks for portfolio construction. Asia Pacific Management Review, 17(2), 113–126.

    Google Scholar 

  • Jobson, J. D., & Korkie, B. (1981). Putting Markowitz theory to work. Journal of Portfolio Management, 7(4), 70–74.

    Article  Google Scholar 

  • Jorion, P. (1986). Bayes-Stein estimation for portfolio analysis. Journal of Financial and Quantitative Analysis, 21(3), 279–292.

    Article  Google Scholar 

  • Levy, H., & Sarnat, M. (1970). International diversification of investment portfolios. American Economic Review, 60, 668–675.

    Google Scholar 

  • Lin, W., Hoffman, P., & Duncan, A. (2009). Global value investing delivers diversification: A multi-strategy perspective. The Journal of Portfolio Management, 35(2), 50–60.

    Article  Google Scholar 

  • Lin, C. M., Huang, J. J., Gen, M., & Tzeng, G. H. (2006). Recurrent neural networks for dynamic portfolio selection. Applied Mathematics and Computers, 175, 1139–1146.

    Google Scholar 

  • Markowitz, H. M. (1952). Portfolio selection. Journal of Finance, 7(1), 77–91.

    Google Scholar 

  • Markowitz, H. M. (1959). Portfolio selection: Efficient diversification of investments. New York: Wiley.

    Google Scholar 

  • Marques, N. C., & Gomes, C. (2010). Maximus-AI: Using Elman neural networks for implementing a SLMR trading strategy. In Y. Bi & M. Williams (Eds.), Knowledge science, engineering and management. KSEM 2010. Lecture notes in computer science (Vol. 6291, pp. 579–584). Berlin, Heidelberg: Springer.

  • Michaud, R. O. (1989). The Markowitz optimization enigma: Is optimized optimal. Financial Analysts Journal, 45, 31–42.

    Article  Google Scholar 

  • Michaud, R. O. (1998). Efficient asset management: A practical guide to stock portfolio optimization and asset allocation. Boston: Harvard Business School Press.

    Google Scholar 

  • Modigliani, F., & Modigliani, L. (1997). Risk-adjusted performance: How to measure it and why. Journal of Portfolio Management, 23(2), 45–54.

    Article  Google Scholar 

  • Neumaier, A., & Schneider, T. (2001). Estimation of parameters and eigenmodes of multivariate autoregressive models. ACM Transactions on Mathematical Software, 27, 27–57.

    Article  Google Scholar 

  • Refenes, A., & Zapranis, A. (1999). Neural model identification, variable selection and model adequacy. Journal of Forecasting, 18, 299–332.

    Article  Google Scholar 

  • Refenes, A. N., Zapranis, A. D., Connor, J. T., & Bunn, D. W. (1994). Neural networks in investment management. Journal of Communication and Finance, 8, 95–101.

    Google Scholar 

  • Schneider, T., & Neumaier, A. (2001). Algorithm 808: ARfit—a matlab package for the estimation of parameters and eigenmodes of multivariate autoregressive models. ACM Transactions on Mathematical Software, 27, 58–65.

    Article  Google Scholar 

  • Shao, J., & Tu, D. (1995). The jackknife and bootstrap. Berlin: Springer.

    Book  Google Scholar 

  • Solnik, B. (1974). Why not diversify internationally rather than domestically? Financial Analysts Journal, 30, 48–54.

    Article  Google Scholar 

  • Treynor, J., & Black, F. (1973). How to use security analysis to improve portfolio selection. Journal of Business, 46, 66–86.

    Article  Google Scholar 

  • Wang, L., Liang, Y., Shi, X., Li, M., & Han, X. (2006). An improved OIF Elman neural networks and its applications to stock market. Knowledge-Based Intelligent Information and Engineering Systems, Lecture Notes in Computer Science, 4251, 21–28.

    Google Scholar 

  • White, H. (1989). Some asymptotic results for learning in single hidden-layer feed-forward network models. Journal of American Statistical Association, 84, 1008–1013.

    Article  Google Scholar 

  • White, H., & Racine, J. S. (2001). Statistical inference, the bootstrap, and neural network modeling with application to foreign exchange rates. IEEE Transactions on Neural Networks, 12, 657–673.

    Article  Google Scholar 

  • Yu, L., Wang, S., & Lai, K. K. (2008). Neural network-based mean–variance skewness model for portfolio selection. Computers & Operations Research, 35, 34–46.

    Article  Google Scholar 

Download references

Author information

Authors and Affiliations

  1. Newhuadu Business School, Minjiang University, No. 200, Xiyuangong Road, Shangjie Town, Minhou County, Fuzhou City, Fujian Province, China

    Hsiao-Fen Hsiao

  2. School of Business, Putian University, No. 1133 Xueyuan Middle Street, Chengxiang District, Putian, 351100, Fujian, China

    Jiang-Chuan Huang

  3. Graduate School of Management, National Yunlin University of Science and Technology, No. 123, University Road, Sec. 3, Douliu City, 640, Yunlin County, Taiwan

    Zheng-Wei Lin

Authors
  1. Hsiao-Fen Hsiao
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Jiang-Chuan Huang
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Zheng-Wei Lin
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Jiang-Chuan Huang.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations

Appendix 1: The Elman neural network

Appendix 1: The Elman neural network

In this study, we adopt the Elman network (implemented by Matlab software); this is a two-layer network with feedback from the first-layer output to the first-layer input, with the recurrent connection allowing the network to both detect and generate time-varying patterns. The two-layer Elman network is specified as follows.

$$ \begin{aligned} a^{1} (t) = \tan sig(IW_{1,1} p + LW_{1,1} a^{1} (t - 1) + b^{1} ) \hfill \\ a^{2} (t) = purelin(LW_{2,1} a^{1} (t - 1) + b^{2} ) \hfill \\ \end{aligned} $$
(9)

where a1 (t), which is the output vector of the recurrent hidden layer, represents the hyper-tangent transference of the input vector, p, and the context unit, LW1,1; tan sig refers to the hyper-tangent activation function; IW1,1 represents the input weight; and b1 is the bias vector.

In the second part of Eq. (9), a2 (t), which is the output vector of the output layer, represents the linear transference of the hidden output vector, a1 (t), and the bias vector, b2; purelin refers to the linear activation function; and LW2,1 represents the hidden weight matrix.

A BFGS quasi-Newton method is used for network training to update the weight and bias values. Given that the network can store information for future reference, it is able to learn temporal patterns as well as spatial patterns.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Hsiao, HF., Huang, JC. & Lin, ZW. Portfolio construction using bootstrapping neural networks: evidence from global stock market. Rev Deriv Res 23, 227–247 (2020). https://doi.org/10.1007/s11147-019-09163-y

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11147-019-09163-y

Keywords

JEL Classification

Search

Navigation