Title: VAR model training using particle swarm optimisation: evidence from macro-finance data

Authors: George Filis, Kyriakos Kentzoglanakis, Christos Floros

Addresses: Department of Economics, University of Portsmouth, Richmond Building, Portland Street, Portsmouth, PO1 3DE, UK. ' School of Computing, University of Portsmouth, Buckingham Building, Lion Terrace, Portsmouth, Hampshire, PO1 3HE, UK. ' Department of Economics, University of Portsmouth, Richmond Building, Portland Street, Portsmouth, PO1 3DE, UK

Abstract: This paper examines the empirical relationship between CPI, oil prices, stock market and unemployment in EU15 using a new computational approach. In particular, we propose a novel approach to train the well-known vector autoregressive (VAR) model using a particle swarm optimisation (PSO) method. Results demonstrate that PSO succeeds in training the model parameters. Furthermore, as the prediction error is found to be low, this strengthens the validity and usability of PSO as a model training method. The empirical results suggest that oil is an important determinant of CPI and stock market changes. Oil price changes affect CPI positively and stock market negatively. Finally, we report no evidence that CPI and unemployment have a negative effect on stock market performance.

Keywords: particle swarm optimisation; PSO; vector autoregressive; VAR model training; macroeconomic indicators; oil prices; stock market performance; unemployment; consumer price index; CPI.

DOI: 10.1504/IJCEE.2009.029150

International Journal of Computational Economics and Econometrics, 2009 Vol.1 No.1, pp.9 - 22

Published online: 06 Nov 2009 *

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