Dynamic Dependence Networks: Financial Time Series Forecasting & Portfolio Decisions

Zoey Yi Zhao, Meng Xie, Mike West
Duke University

May 14 2015

Published version: Applied Stochastic Models in Business & Industry, 32:311-339


We discuss Bayesian forecasting of increasingly high-dimensional time series, a key area of application of stochastic dynamic models in the financial industry and allied areas of business. Novel state-space models characterizing sparse patterns of dependence among multiple time series extend existing multivariate volatility models to enable scaling to higher numbers of individual time series. The theory of these {\em dynamic dependence network} models shows how the individual series can be decoupled for sequential analysis, and then {\em recoupled} for applied forecasting and decision analysis. Decoupling allows fast, efficient analysis of each of the series in individual univariate models that are linked-- for later recoupling-- through a theoretical multivariate volatility structure defined by a sparse underlying graphical model. Computational advances are especially significant in connection with model uncertainty about the sparsity patterns among series that define this graphical model; Bayesian model averaging using discounting of historical information builds substantially on this computational advance. An extensive, detailed case study showcases the use of these models, and the improvements in forecasting and financial portfolio investment decisions that are achievable. Using a long series of daily international currency, stock indices and commodity prices, the case study includes evaluations of multi-day forecasts and Bayesian portfolio analysis with a variety of practical utility functions, as well as comparisons against commodity trading advisor benchmarks.

Keywords: Bayesian forecasting; discount model averaging; dynamic graphical model; graphical model uncertainty; multiregression dynamic model; portfolio optimization; sparse dynamics

Acknowledgements: This work was completed while Zoey Zhao was a PhD student in the Department of Statistical Science at Duke University, and aspects of the work were developed by Meng Xie while an undergraduate student at Duke. The research was partly supported by a grant from the National Science Foundation [DMS-1106516]. Any opinions, findings and conclusions or recommendations expressed in this work are those of the authors and do not necessarily reflect the views of the NSF.


Bayesian forecasting, Cholesky stochastic volatility, Dynamic graphical model, Model averaging, Parallel computation, Portfolio optimization, Sparse parental sets


PDF icon 2014-03_1.pdf

BibTeX Citation: 

  author = {Z. Y. Zhao and M. Xie and M. West},
  title = {Dynamic dependence networks: {F}inancial time series forecasting \& portfolio decisions (with discussion)},
  journal = {Applied Stochastic Models in Business and Industry},
  year = {2016},
  volume = {32},
  pages = {311-339},
  note = {First published online: March 25, 2016},
  doi = {10.1002/asmb.2161}