Prediction of Stock Returns Using Classical and Intelligent Techniques: Evidence from BSE Sensex
International Journal of Applied Business and Economic Research, Vol. 4, No. 2, pp. 109-123, 2006
16 Pages Posted: 31 Oct 2010 Last revised: 14 Jan 2015
Date Written: 2006
This paper is an attempt to predict stock returns using classical (AR) and intelligent (ANN) techniques. AR and ANN techniques are also used to test the efficient market hypotheses using long time-series of daily data of BSE Sensex for the period of January 1997 to September 2005. An attempt has also been made to compare the predictive power of autoregressive (AR) model and artificial neural network (ANN). The present study shows that to a large extent stock morket returns are predictable. Profitable investment decisions may be taken using linear and nonlinear techniques of prediction. Prediction improves by using ANN over AR model. Investment decisions based ANN prediction performs better compared to AR Prediction. It is observed from the obtained result that MSE and RMSE decrease as the number of observations in training set increases. This is equally applicable to both linear and non-linear (ANNN) modeling. The present study does not support efficient market hypotheses.
Keywords: Efficient Market hypotheses, BDS Test, Artificial Neural Network, Backpropogation
JEL Classification: G12, G14, C45, C53
Suggested Citation: Suggested Citation