An intelligent system for forex trading: Hybrid ANN with GARCH and intrinsic mode functions

Time series forecasting can be viewed as an indispensable topic under discussion in a wide range of fields including, medicine, engineering, finance and economics, etc. As assumed by the conventional time series models static environmental conditions are rare to exist in real life situations. This study addressed this issue in case of financial markets by proposing an intelligent decision support system for Forex (Foreign Exchange) trading which can be used even under dynamic environmental conditions. The study proposes to identify clusters of time points relevant to dynamic environments defined by the release of scheduled news items. Then the proposed system utilizes the relevant database of time points to derive the forecast. The proposed intelligent system is supported with a hybrid neural network approach which combines generalized autoregressive conditional heteroskedasticity (GARCH) volatility estimates and intrinsic mode functions derived from Empirical Mode Decomposition. The incorporation of clustering of time points found significance in the forecasting process utilized in the proposed intelligent system over the method which ignores clustering. Moreover, the proposed system is efficient enough in delivering the trading decision yielding instant profitable opportunities, 57% of times and aggregated profit over hundred trading positions remained positive at almost all the time points.