Title: Predicting the present with Bayesian structural time series

Authors: Steven L. Scott; Hal R. Varian

Addresses: Google, Inc., 1600 Amphitheatre Parkway, Mountain View, CA 94043, USA ' Google, Inc., 1600 Amphitheatre Parkway, Mountain View, CA 94043, USA

Abstract: This article describes a system for short term forecasting based on an ensemble prediction that averages over different combinations of predictors. The system combines a structural time series model for the target series with a regression component capturing the contributions of contemporaneous search query data. A spike-and-slab prior on the regression coefficients induces sparsity, dramatically reducing the size of the regression problem. Our system averages over potential contributions from a very large set of models and gives easily digested reports of which coefficients are likely to be important. We illustrate with applications to initial claims for unemployment benefits and to retail sales. Although our exposition focuses on using search engine data to forecast economic time series, the underlying statistical methods can be applied to more general short term forecasting with large numbers of contemporaneous predictors.

Keywords: Bayesian model averaging; Bayesian structural time series models; Markov chain Monte Carlo; MCMC; economic time series; machine learning; predicting the present; spike and slab priors; state space models; short term forecasting; ensemble prediction; unemployment benefits; retail sales.

DOI: 10.1504/IJMMNO.2014.059942

International Journal of Mathematical Modelling and Numerical Optimisation, 2014 Vol.5 No.1/2, pp.4 - 23

Received: 08 May 2021
Accepted: 12 May 2021

Published online: 20 Mar 2014 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article