M Thenmozhi

This study investigates the performance of four training algorithms, namely, Standard Backpropagation (SBP), Scaled Conjugate Gradient (SCG), Resilient Backpropgation (RBP) and Levenberg-Marquardt (LM) Backpropagation in forecasting three financial time series, namely, Indian call rates, INR/USD exchange rates and S&P CNX Nifty Index. The models are trained from historical data using six technical indicators. The predicted results show that among the four training algorithms, LM based model outperforms other models when measured on commonly used non-penalty based metrics while SCG based model outperforms the other models when direction and sign based performance metrics are used.

In this study forecasting of the NIFTY stock index futures returns is carried out using backpropagation and recurrent neural network model and a linear ARIMA model. A comparison of different models shows that for NIFTY index futures returns backpropagation neural network model outperforms the recurrent neural network and the traditional ARIMA models. Moreover recurrent neural network models outperform the traditional ARIMA models. A 3-2-1 neural network architecture is best fit for forecasting futures returns.