Abstract
This chapter explores the possibility to leverage semantic knowledge for robust estimation of correlations among financial assets. A graphical model for high-dimensional stochastic dependence termed a “vine” structure, which is derived from copula theory, is introduced here. To model the prior semantic knowledge, we use a neural network-based language model to generate distributed semantic representations for financial documents. The semantic representations are used for computing similarities between the assets they respectively refer. The constructed dependence structure is experimented with real-world data. Results suggest that our semantic vine construction-based method is superior to the state-of-the-art covariance matrix estimation method, which is based on an arbitrary vine that at least guarantees robustness of the estimated covariance matrix. The effectiveness of using semantic vines for robust correlation estimation for Markowitz’s asset allocation model on a large scale of assets (up to 50 stocks) is also showed and discussed.
We use a machine, or the drawing of a machine, to symbolize a particular action of the machine. — Ludwig Wittgenstein
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Shallow refers to the neural networks that only have one hidden layer of neurons.
- 2.
The correlation between a 1 and a 3 conditions on a pivot asset a 2. Therefore, we use a different type of dashed link to denote this conditional correlation. The dashed link is abbreviated as 1, 3|2.
- 3.
Proof of this theorem uses trigonometric substitution. For details, see Lemma 12 in Bedford and Cooke [11].
- 4.
This is defined as the optimal truncation of vines as a minimum number of edges would have large absolute partial correlations and rest of the edges are assumed insignificant (independent).
- 5.
Information on the stock list is elaborated in Appendix A.
- 6.
- 7.
- 8.
Retrieved from the Internet on 2017-10-09.
- 9.
See Sect. 4.2.3 for the definition of the optimal vine truncation.
- 10.
This time span is roughly chosen because it is reasonable to assume the asset correlations keep the same. If simulation is carried out for a longer period, we have to access historical corpus of the Reuters Company Business Descriptions and Wikipedia pages, which is out of scope for our discussion.
References
K. Aas, D. Berg, Models for construction of multivariate dependence – a comparison study. Eur. J. Financ. 15, 639–659 (2009)
H. Bai, F.Z. Xing, E. Cambria, W.-B. Huang, Business taxonomy construction using concept-level hierarchical clustering, in The First Workshop on Financial Technology and Natural Language Processing (FinNLP-IJCAI), 2019, pp. 1–7
T. Bedford, R.M. Cooke, Probability density decomposition for conditionally dependent random variables modeled by vines. Ann. Math. Artif. Intell. 32, 245–268 (2001)
T. Bedford, R.M. Cooke, Vines: a new graphical model for dependent random variables. Ann. Stat. 30(4), 1031–1068 (2002)
Y. Bengio, R. Ducharme, P. Vincent, A neural probabilistic language model. J. Mach. Learn. Res. 3, 1137–1155 (2003)
L.K.C. Chan, J. Lakonishok, B. Swaminathan, Industry classification and return comovement. Financ. Anal. J. 63(6), 56–70 (2007)
I. Chaturvedi, Y.-S. Ong, I. Tsang, R.E. Welsch, E. Cambria, Learning word dependencies in text by means of a deep recurrent belief network. Knowl. Based Syst. 108, 144–154 (2016)
R.M. Cooke, D. Kurowicka, K. Wilson, Sampling, conditionalizing, counting, merging, searching regular vines. J. Multivar. Anal. 138, 4–18 (2015)
W. Croft, D.A. Cruse, Cognitive Linguistics (Cambridge University Press, New York, 2004)
A.B. Davidow, J.D. Peterson, A modern approach to asset allocation and portfolio construction. Technical Report MKT81752HL-02, Schwab Center for Financial Research, 2014
F. Durante, C. Sempi, Principles of Copula Theory (CRC Press, Boca Raton, 2016)
G. Elidan, Copulas in machine learning, in Copulae in Mathematical and Quantitative Finance, vol. 213 (Springer, Berlin/Heidelberg, 2013), pp. 39–60
E.F. Fama, K.R. French, Luck versus skill in the cross-section of mutual fund returns. J. Financ. 65(5), 1915–1947 (2010)
K.K. Hung, C.C. Cheung, L. Xu, New Sharpe-ratio-related methods for portfolio selection, in Proceedings of the Conference on Computational Intelligence for Financial Engineering (CIFEr), 2000, pp. 34–37
D.P. Kingma, J. Ba, Adam: a method for stochastic optimization, in Proceedings of International Conference on Learning Representations, 2015
D. Kurowicka, H. Joe (eds.), Dependence Modeling: Vine Copula Handbook (World Scientific, London, 2011)
Q. Le, T. Mikolov, Distributed representations of sentences and documents, in Proceedings of the 31st International Conference on Machine Learning (ICML), 2014, pp. 1188–1196
G. Leech, Semantics: The Study of Meaning, 2 edn. (Harmondsworth, Penguin, 1981)
X. Li, H. Xie, Y. Song, S. Zhu, Q. Li, F.L. Wang, Does summarization help stock prediction? A news impact analysis. IEEE Intell. Syst. 30(3), 26–34 (2015)
L. Luo, Y. Xiong, Y. Liu, X. Sun, Adaptive gradient methods with dynamic bound of learning rate, in Proceedings of International Conference on Learning Representations, 2019
R.C. Merton, On estimating the expected return on the market: an exploratory investigation. J. Financ. Econ. 8(4), 323–361 (1980)
D. Metzler, W.B. Croft, A Markov random field model for term dependencies, in Proceedings of the 28th Annual International Conference on Research and Development in Information Retrieval (SIGIR), 2005, pp. 472–479
T. Mikolov, I. Sutskever, K. Chen, G. Corrado, J. Dean, Distributed representations of words and phrases and their compositionality, in Proceedings of the 26th International Conference on Neural Information Processing Systems (NIPS), 2013, pp. 3111–3119
N.M. Neykov, P. Filzmoser, P.N. Neytchev, Robust joint modeling of mean and dispersion through trimming. Comput. Stat. Data Anal. 56(1), 34–48 (2012)
A. Panagiotelis, C. Czado, H. Joe, J. Stoeber, Model selection for discrete regular vine copulas. Comput. Stat. Data Anal. 106, 138–152 (2017)
H. Qiu, F. Han, H. Liu, B. Caffo, Robust portfolio optimization, in Neural Information Processing Systems (NIPS), 2015, pp. 46–54
S.T. Rachev, S.V. Stoyanov, A. Biglova, F.J. Fabozzi, An empirical examination of daily stock return distributions for U.S. Stocks, in Data Analysis and Decision Support (Springer, Berlin/Heidelberg, 2005), pp. 269–281
D. Tran, D.M. Blei, E.M. Airoldi, Copula variational inference, in Advances in Neural Information Processing Systems (NIPS) (Springer, Cham, 2015), pp. 3564–3572
R.R. Trippi, J.K. Lee, Artificial Intelligence in Finance & Investing (Irwin Professional Publishing, Chicago, 1996)
R.E. Welsch, X. Zhou, Application of robust statistics to asset allocation models. Revstat Stat. J. 5(1), 97–114 (2007)
F.Z. Xing, E. Cambria, R.E. Welsch, Growing semantic vines for robust asset allocation. Knowl. Based Syst. 165, 297–305 (2019)
L. Zhang, C. Aggarwal, G.-J. Qi, Stock price prediction via discovering multi-frequency trading patterns, in The 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2017, pp. 2141–2149
Z. Zhu, R.E. Welsch, Robust dependence modeling for high-dimensional covariance matrices with financial applications. Ann. Appl. Stat. 12(2), 1228–1249 (2018)
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2019 Springer Nature Switzerland AG
About this chapter
Cite this chapter
Xing, F., Cambria, E., Welsch, R. (2019). Computational Semantics for Asset Correlations. In: Intelligent Asset Management. Socio-Affective Computing, vol 9. Springer, Cham. https://doi.org/10.1007/978-3-030-30263-4_4
Download citation
DOI: https://doi.org/10.1007/978-3-030-30263-4_4
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-30262-7
Online ISBN: 978-3-030-30263-4
eBook Packages: Biomedical and Life SciencesBiomedical and Life Sciences (R0)