Multistep forecast of FX rates using an extended self-organizing regressive neural network

Abstract

In this paper, an extended self-organizing regressive neural network is proposed for multistep forecasting of time series. The main features of this method are building input segments with various lengths for training and the network capable of learning multiple regressive models for forecasting various horizons. The inter dependencies among future points are preserved and this results in all forecasting tasks naturally. Experiments on foreign exchange rates show that the new method significantly improves the performance in multistep forecasting compared to the existing methods.