Online Learning and Variable Selection for High-Dimensional Time Series with Simultaneous Graphical Dynamic Linear Models

• Activity Number: 478 - Scalable Bayesian Models for Time Series and Dynamic Networks: Making an Impact in Business and Socio-Economic Applications
• Type: Invited
• Date/Time: Wednesday, July 31, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Lutz F Gruber* and Mike West
• Companies: QuantCo, Inc. and Duke University

Abstract

Simultaneous graphical dynamic linear models (SGDLMs) define an ability to scale on-line Bayesian analysis and multivariate volatility forecasting to higher-dimensional time series. Advances in the methodology of SGDLMs involve a novel, adaptive method of simultaneous predictor selection in forward filtering for on-line learning and forecasting. This Bayesian methodology for dynamic variable selection and Bayesian computation for scalability are highlighted in a case study evidencing the potential for improved short-term forecasting of large-scale volatility matrices. In financial forecasting and portfolio optimization with a 400-dimensional series of daily stock prices, analysis demonstrates SGDLM forecasts of volatilities and co-volatilities that contribute to quantitative investment strategies to improve portfolio returns. Performance metrics linked to the sequential Bayesian filtering analysis define a leading indicator of increased financial market stresses, comparable to but leading standard financial risk measures. Parallel computation using GPU implementations substantially advance the ability to fit and use these models.