Forecasting financial multivariate time series with neural networks

Abstract

An integrated approach for modelling the behaviour of financial markets with artificial neural networks (ANNs) is presented. The method allows to forecast financial time series. Its originality lies in the fact that it is based on statistics and macroeconomics principles and it integrates fundamental economic knowledge in a multivariate nonlinear time series ANN model. The core of the work is a feasibility analysis. This is seldom attempted in ANN work and consists in a series of different univariate and multivariate, linear and nonlinear statistical tests. Here we use aggregated input indicators as a new pre-processing step. The feasibility analysis evaluate "a priori" chance of forecasting the defined system and help to define the topology of the ANN. The method is applied to a real-life case study, the Swiss bond interest rate forecasting. Results giving out-of-sample performance are discussed.