Choice of Methods in Predicting Index Return Returns Using Artificial Neural Network for National Stock Exchange of India
Abstract
This study uses artificial neural network (ANN) in forecasting the NIFTY. The paper trains several ANN by using back propagation algorithm and they were assessed. The study uses long range data for a period of 1st April 2005 to 31st closing daily return. A total of 2485 data points were trained. The study shows that algorithm provides 99 percentage accuracy. The study is taken as an alternative to the liner models that are mostly used in prediction of the returns. The best prediction was found through OSS-TANSIG which had a validity of 0.9407. With respect to its usability as given by the weighted totals most of the models are of same worth. Hence, the choice of model in real life, especially for very long period data is insignificant and not of high order. This scanty difference in method may have arise due to very long duration data, since BPNN works best in very short and tick data format.
