Forecasting the stock price movement of a company and stock index is a classic problem. Efficient Market Hypothesis clearly asserts that it is not possible to exactly predict the stock prices of corporate entities, due to the existence of random walk behavior in stock markets (Fama, 1970). The movements of stock prices and stock indices are influenced by many macroeconomic variables such as political events, policies of the corporate enterprises, general economic conditions, commodity price index, bank rate, loan rates, foreign exchange rates, investors' expectations, investors' choices and the human psychology of stock market investors (Miao et al., 2007). Neural networks are a class of generalized, non-linear and non-parametric models, inspired by the studies of human brain. The feed-forward networks are the most widely used (Ou & Wang, 2009). Prediction of stock market movements has become increasingly difficult due to uncertainties, involved with the probable future outcomes. At a particular point of time, there could be trends, cycles and random walk or a combination of three cases/events (Robert & David, 2011). Closing price of a stock/index has been used, as one of the important statistical data, to derive useful information about the current and probable future movement pattern of stock market (Defu et al., 2005). In data deterministic approach, a layer is employed to convert each of the technical indicator's continuous value, from +1 to -1, indicating the probable future growth/decline movements. This layer explains the manner of stock market movements, on both upwards and downwards direction, across the time periods (Shuai & Wei, 2014). The data deterministic approach could forecast the future trend of stock market and to provide stock information signs, for taking better investment decision of buying and selling of stocks by the investors (Jigar et al., 2015a).
REVIEW OF LITERATURE
An extensive review of literature, in the area of prediction of stock indices, has been done by the researchers. Prediction of stock price movements of individual stocks and stock market indices is explained by Amitai (1976), who also explained the difficulties involved in making specific and accurate forecasting of financial markets. Wang & Leu (1996) forecasted six week stock price trend, based on past four years stock price movements of Taiwan stock market, by using recurrent neural network. Hansen & Nelson (2003) applied a time-delay neural network to predict the stock price movements and the results of future trend prediction, using the hybrid system, proved to be promising. Kim (2003) used twelve technical indicators, to make prediction of daily stock price changes and stock index movements of Korea Composite Stock Price Index (KOSPI). Simulation results of Shanghai Composite Index show that neural networks could be applied to maximize the returns of stock market investment (Defu et al., 2005). Franklin Kuo classified the networks into linear, passive, reciprocal, causal and time invariant and each network approaches has unique and different inherent characteristic properties accordingly. Teo & Douglas (2006) calculated the profitability of stock indices, based on daily trades of S&P 500 (India), DAX (Germany), Topix (Japan) and FTSE (UK). Andre & Beatriz (2008) adopted neural network, to forecast the stock market returns of emerging economies. Ou & Wang (2009) tried out ten different data mining techniques, in order to predict the stock price movements of Hang Seng index of Hong Kong stock market. According to Robin & Ryan (2009), novelty, complexity and anonymity of forecasting the stock markets, influenced the observers to have reservations about the prediction outcome of stock price movements. Nair et al. (2011) predicted the closing value of next day, for five international stock indices, using an adaptive artificial neural network system. Garg et al. (2013) analyzed the model selection criteria, for two data transmission models of the stocks, indexed in New York Stock Exchange. Jigar et al. (2015a) used regression method to predict the trends. Prediction performance of BSE- Sensex, NSE-Nifty, Reliance Industries and Infosys Limited were measured and compared in a group of stock indices and individual stocks respectively. Shin & Shie (2015) used the Box-Jenkins model, spectral analysis and Markov process, to forecast the stock prices. Based on the above reviews, the researchers applied one of the neural network methods i.e., data deterministic approach is used in this study, to predict the robust growing Indian stock market.
STATEMENT OF THE PROBLEM
Prediction of the movements of financial markets is one of the classic issues. Profit making in stock market investments is, linked to the level of financial literacy, financial intelligence and financial investment practice. Less exposure to these aspects, among the financial investors and absence of proven forecasting techniques, to exactly predict the probable futuristic movement pattern of stock price/index values, cause the magnitude and severity of this issue which also vindicates this kind of study (Melek & Derya, 2010).
NEED OF THE STUDY
This study would help the investors, ranging from domestic retail investors, financial institutions, mutual funds, investment banks, to foreign institutional investors, to take timely and well-informed investment decisions, based on scientific thinking and rational approach (Amitai, 1976). Availability of alternate investment options, absence of prudent prediction methods and incidence of lower level of financial literacy reiterate this kind of study.
The primary objective is to find out the existing stock index movement pattern and to predict the probable future movements of BSE-Sensex and NSE-Nifty.
[NH.sub.1]: There is no corresponding relationship between the upward/downward movements of BSE-Sensex and NSE-Nifty, during the pre-global financial crisis period.
[NH.sub.2]: There is no corresponding relationship between the upward/downward movements of BSE-Sensex and NSE-Nifty, during the post-global financial crisis period.
Sampling Design of the Study
Based on the free-float market capitalization, as on 07/02/2017, S&P BSE-Sensex (Rs. 53, 88, 277 Crores) and CNX NSE-Nifty (Rs. 49, 27, 183 Crores) were considered, since these two stock indices signify the overall direction of the stock market movements in India.
Sources of Data
The secondary data of the daily closing stock index values of BSE-Sensex and NSE-Nifty were collected from respective websites of Bombay Stock Exchange (www.bseindia.com) and the National Stock Exchange of India (www.nseindia.com).
Period of the Study
The study focused on the behavior of stock price/index movements of eighteen years from 01st January 1999 to 31st December 2016, before and after the global financial crisis.
Statistical Tools Used in the Study
To forecast the movements of BSE-Sensex and NSE-Nifty, ten technical indicators (Simple Moving Average, Weighted Moving Average, Momentum, Stochastic K%, Stochastic D%, Relative Strength Index, Moving Average Convergence and Divergence, Larry Williams R, A/D Oscillator and Commodity Channel Index) were used.
Limitations of the Study
Each technical indicator has its own optimization about the stock price movements. Only two stock indices, BSE-Sensex and NSE-Nifty, were used as sample.
DATA ANALYSIS AND INTERPRETATION
The daily closing values of stock indices of BSE-Sensex and NSE-Nifty, which were collected and analyzed for this study, were as follows:
The Observations of the Total Dataset
Observation of BSE-Sensex
Table 1 exhibits the observations of the total dataset (continuous), relating to BSE-Sensex, during the study period from 1999 to 2016. The analysis of pre-crisis period (i.e., 1999 to 2007) shows that in 1999, the number of bull movements was recorded as 136, which represented 54.83% while there were 112 bear movements, which accounted for 45.17%, out of total stock index movements of 248. In the...