Title: Bayesian emulation for optimization: Multi-step portfolio decision analysis
Authors: Kaoru Irie - University of Tokyo (Japan) [presenting]
Mike West - Duke University (United States)
Abstract: Bayesian analysis is discussed for portfolio studies involving multi-step forecasts and decisions in financial time series. Using classes of economically and psychologically relevant multi-step ahead utility functions, we develop solutions to the resulting Bayesian expected utility optimization problem. The solution paths involve mapping the technical structure of (some) optimization problems to those of parallel, synthetic Bayesian inference problems in ``emulating'' statistical models. This provides access to standard Bayesian simulation and optimization methods that then yield indirect solutions of the decision problems. Study of sequential portfolio studies with multivariate currency, commodity and stock index time series illustrate the approach and show some of the practical utility and benefits of the new Bayesian decision analysis framework and methodology.