Forecasting foreign exchange rates using hybrid functional link RBF neural network and Levenberg-Marquardt learning algorithm

Abstract

This paper proposes a novel nonlinear ensemble forecasting model integrating functional link (FL) with radial basis function (RBF) neural network in order to improve prediction performance. In addition to the traditional parameters like the centers, widths and output weights, the input weights of the connections between the input and hidden layer are also adjusted during the training process. The developed algorithm is introduced for designing a compact FLRBF (Functional link Radial Basis Function) network and performing efficient training process. A certain set of prominent trading indicators, together with the moving average convergence/divergence and relative strength index, are also utilized in the anticipated model. The proposed approach is applied to currency exchange prediction to test the main properties of FLRBF network, including its generalization ability, tolerance to input noise, and online learning ability. More specifically, the trading and statistical performance of all models are investigated in a forecast simulation of the exchange rates between American Dollar and four other major currencies, Euro, Indian Rupee, Canadian Dollar, Australian Dollar, etc. over the period January 2004 to January 2014 using the last years for out-of-sample testing. Further the FLRBF network prediction performance is also compared with linear, nonlinear and hybrid neural networks. As it turns out, the FLRBF architecture outperforms all other models in terms of statistical accuracy and trading efficiency for the three exchange rates. Different performance indicators such as MAE (Mean Absolute Error), RMSE (Root Mean Squared Error), and MAPE(Mean Absolute Percentage Error) are employed in order to evaluate the performance of the proposed model.