Artificial Neural Network (ANN) models have been proved to be powerful predictive tools, where a variable is explained by a set of explanatory variables without assuming any structural or linear relationship among the variables. In the field of finance, a large number of models, especially those derived from the field of econometrics, are used for forecasting stock returns. This paper intends to test the forecasting ability of ANN models using Nifty index data from Indian stock market. Daily time series data of the index of National Stock Exchange is analyzed using a three-layer architecture of the ANN. The results of the study reveal that ANN models could efficiently predict daily returns of Nifty index for a given period under investigation. The results of this study are significant value addition to the trading decisions in the stock index futures with special reference to Indian stock market.

Forecasting the future returns has always been a major concern for the players in the stock markets. Since any stock market comprises a variety of players including investors, they all tend to employ different strategies to gain the best out of the market. Long-term investors perform a thorough analysis of the fundamentals of the investment opportunities that they consider, while short-term investors, known as traders, tend to base their trading decisions on the forecasts of near future returns from the specific stocks. As stock markets are highly unpredictable and volatile, returns in the short run also become quite unpredictable, thereby making the trading decisions highly risky for the short-term traders. Returns in the stock markets move in tandem with market participants' sentiments and a host of other factors, and the short-term traders find it very difficult in quantifying all those factors, to reap benefits from each market move.

Several studies have examined the index forecasting in the context of intra-stock market as well as inter-stock markets, with the help of various models including those involving econometrics principles and those using neural network . But the problem still persists with the forecasting of index returns. The classical rule of demand and supply in economics can be considered as an alternative approach. It is believed that the price of a security at any given point of time is the result of the equilibrium of the demand and supply raised by the different market players. In the stock market context, this demand and supply made by market players is totally driven by their sentiments which is influenced by several known and unknown factors. So, if we count on the price of the securities at different point of time reflecting different relational structures, then we are actually discounting all the price-impacting factors in the market (Roy and Roy, 2008).

Computational Mathematics Journal, Indian Stock Market, National Stock Exchange, Artificial Neural Network Models, Modeling Procedures, Neural Network Models, Efficient Market Hypothesis, Saudi Arabian Companies, Australian Companies, Genetic Algorithm, Multiple Discriminant Analysis, Biological Neural Networks.