Forecasting and trading the EUR/USD exchange rate with stochastic Neural Network combination and time-varying leverage

The motivation of this paper is to investigate the use of a Neural Network (NN) architecture, the Psi Sigma Neural Network (PSN), when applied to the task of forecasting and trading the Euro/Dollar (EUR/USD) exchange rate using the European Central Bank (ECB) fixing series and to explore the utility of Kalman Filters in combining NN forecasts. This is done by benchmarking the statistical and trading performance of PSN with a Naive Strategy, an Autoregressive Moving Average (ARMA) model and two different NN architectures, a Multi-Layer Perceptron (MLP) and a Recurrent Network (RNN). We combine our NN forecasts with Kalman Filter, a traditional Simple Average, the Bayesian Average, the Granger–Ramanathan's Regression Approach (GRR) and the Least Absolute Shrinkage and Selection Operator (LASSO). Finally, we apply a time-varying leverage strategy based on RiskMetrics volatility forecasts in order to further improve the forecasting performance of our models and combinations. The statistical and trading performance of our models is estimated throughout the period of 2002–2010, using the last two years for out-of-sample testing. In terms of our results, the PSN outperforms all models' individual performances in terms of statistical accuracy and trading performance. The forecast combinations also present improved empirical evidence, with Kalman Filters outperforming by far its benchmarks. We also note that after the application of the time varying leverage, all models except ARMA show a substantial increase in their trading performance.

► We investigate the use of the Psi Sigma Neural Network (PSN) in forecasting.
► We explore the utility of Kalman Filters in combining Neural Networks.
► We introduce a time-varying leverage strategy.
► The PSN outperforms all models' individual performances.
► Kalman Filters outperform its benchmark forecast combination techniques.