Exchangeable models of financial correlations matrices. Bayesian nonparametric models and network derived measures of financial assets


De Finetti theorem establishes the conceptual basis of Bayesian inference replacing the independent and identically distributed sampling hypothesis prevalent in frequentist statistics with the much easier to justify in practical settings hypothesis of exchangeability. In this paper we make use of the extension of the concept of exchangeability from sequences to arrays arguing that the invariance to ordering is a much more tenable assumption than independent and identically distributed sampling in the financial modeling problems. Making use of the celebrated Aldous-Hoover representation theorem of exchangeable matrix we construct a Bayesian non-parametric model of the financial returns correlation matrices arguing that a Bayesian approach can mitigate many of the known shortcomings of the usual Pearson correlation coefficient. We posit the correlation matrix to be an exchangeable matrix and construct a Bayesian neural network to estimate the functions from the Aldous-Hoover representation theorem. The correlation matrix model is coupled with a Student-t likelihood (accounting for the heavy tails of financial returns). The model is estimated with a Hamiltonian Monte Carlo sampler. The samples are used to construct an ensemble of networks where each edge is weighted by the size of the correlation between two financial instruments. Various centrality measures are being calculated (betweenness, eigenvector) for each network of the ensemble allowing us to obtain a probabilistic view of each financial instrument’s importance. We also construct a minimum spanning tree associated with the mean correlation matrix allowing us to visualize the most important financial instruments from the universe selected.

If the inline PDF is not rendering correctly, you can download the PDF file here.

  • Aldous, D. J. (1981). Representations for partially exchangeable arrays of random variables. Journal of Multivariate Analysis, 11(4), 581-598.

  • Cont, R. (2001). Empirical properties of asset returns: stylized facts and statistical issues.

  • De Finetti, B. (2017). Theory of probability: A critical introductory treatment (Vol. 6). John Wiley & Sons.Aldag, R.J., & Stearns, T.M. (1991). Management (2nd ed.). Cincinnati, OH: South-Western Publishing.

  • Fiedor, P. (2014). Networks in financial markets based on the mutual information rate. Physical Review E, 89(5), 052801.

  • Hoover, D. N. (1979). Relations on probability spaces and arrays of random variables. Preprint, Institute for Advanced Study, Princeton, NJ, 2.’

  • Hoffman, M. D., & Gelman, A. (2014). The No-U-Turn sampler: adaptively setting path lengths in Hamiltonian Monte Carlo. Journal of Machine Learning Research, 15(1), 1593-1623.

  • Hornik, K., Stinchcombe, M., & White, H. (1989). Multilayer feedforward networks are universal approximators. Neural networks, 2(5), 359-366.

  • Kenett, D. Y., Huang, X., Vodenska, I., Havlin, S., & Stanley, H. E. (2015). Partial correlation analysis: Applications for financial markets. Quantitative Finance, 15(4), 569-578.

  • Kumar, S., & Deo, N. (2012). Correlation and network analysis of global financial indices. Physical Review E, 86(2), 026101.

  • Laloux, L., Cizeau, P., Bouchaud, J. P., & Potters, M. (1999). Noise dressing of financial correlation matrices. Physical review letters, 83(7), 1467.

  • Ledoux, M. (2001). The concentration of measure phenomenon (No. 89). American Mathematical Soc..

  • Lloyd, J., Orbanz, P., Ghahramani, Z., & Roy, D. M. (2012). Random function priors for exchangeable arrays with applications to graphs and relational data. In Advances in Neural Information Processing Systems (pp. 998-1006).

  • Mantegna, R. N. (1999). Hierarchical structure in financial markets. The European Physical Journal B-Condensed Matter and Complex Systems, 11(1), 193-197.

  • Miller, K., Jordan, M. I., & Griffiths, T. L. (2009). Nonparametric latent feature models for link prediction. In Advances in neural information processing systems (pp. 1276-1284).

  • Namaki, A., Shirazi, A. H., Raei, R., & Jafari, G. R. (2011). Network analysis of a financial market based on genuine correlation and threshold method. Physica A: Statistical Mechanics and its Applications, 390(21-22), 3835-3841.

  • Neal, R. M. (2012). Bayesian learning for neural networks (Vol. 118). Springer Science & Business Media.

  • Newman, M. (2010). Networks: an introduction. Oxford university press.

  • Onnela, J. P., Kaski, K., & Kertész, J. (2004). Clustering and information in correlation based financial networks. The European Physical Journal B, 38(2), 353-362.

  • Onnela, Jukka-Pekka, Anirban Chakraborti, Kimmo Kaski, Janos Kertesz, and Antti Kanto. “Asset trees and asset graphs in financial markets.” Physica Scripta 2003, no. T106 (2003): 48.

  • Orbanz, P., & Roy, D. M. (2015). Bayesian models of graphs, arrays and other exchangeable random structures. IEEE transactions on pattern analysis and machine intelligence, 37(2), 437-461.

  • Plerou, V., Gopikrishnan, P., Rosenow, B., Amaral, L. A. N., & Stanley, H. E. (1999). Universal and nonuniversal properties of cross correlations in financial time series. Physical review letters, 83(7), 1471.

  • Peiro, A. (1999). Skewness in financial returns. Journal of Banking & Finance, 23(6), 847-862.

  • Sengupta, A. M., & Mitra, P. P. (1999). Distributions of singular values for some random matrices. Physical Review E, 60(3), 3389.

  • Tumminello, M., Lillo, F., & Mantegna, R. N. (2010). Correlation, hierarchies, and networks in financial markets. Journal of Economic Behavior & Organization, 75(1), 40-58.

  • Wolfe, Patrick J., and Sofia C. Olhede. “Nonparametric graphon estimation.” arXiv preprint arXiv:1309.5936 (2013)


Journal + Issues