A Generalized Linear and Nonlinear (GLN) model for complex financial time series forecasting

Time series forecasting is one of the most important and well- known tools that can be applied to a wide range of decision making problems with a high degree of accuracy. Many decision makers in the finance need to forecast financial prices in order to make financial decisions. However, literature indicates that forecasting of the financial markets is ever not a simple task. Several researchers believe that the most important reason of the financial markets unpredictability is the complexity and nonlinearity of the financial data. Artificial neural networks (ANNs) are flexible computing frameworks and universal approximators, which can accurately be applied to nonlinear data sets. However, using neural networks to model linear problems have yielded mixed results, and hence; it is not wise to apply ANNs blindly to any type of data. It is the main reason of proposing the hybrid linear/nonlinear methodologies in the literature of financial forecasting. In this paper, a new generalized linear and nonlinear (GLN) model is proposed in order to combine the linear and nonlinear models. In this way, the auto-regressive integrated moving average (ARIMA) and multi-layer perceptrons (MLPs) are chosen as linear and nonlinear models, respectively. Empirical results of the British pound against the US dollar exchange rate indicate that the proposed model can be an effective way to improve forecasting accuracy achieved by traditional hybrid models and also either of the components used separately.