Comparison of different NN training functions of NARX architecture for financial time series

Neural networks are widely applicable in computer science, machine learning and other research area. It is also used in the field of financial time series analysis. In this study we have done comparative analysis of various training functions of neural network using nonlinear autoregressive with exogenous output architecture. This study provides some intuition about which is the best algorithm to train a neural network for financial time series. And which algorithms is best to forecast multi-step ahead and one-step ahead values of financial time series.