An experimental analysis of forecasting the high frequency data of matured and emerging economies stock index using data mining techniques

Abstract

In this paper we study the applicability of three data mining techniques, viz. backpropagation neural network (BPNN), support vector regression (SVR) and multivariate adaptive regression splines (MARSplines) for one step ahead forecast of stock index using intraday data. High frequency data of S&P CNX Nifty (Nifty) from India and Nasdaq Composite Index (NCI) from the USA are used for the analysis to check the efficiency of these techniques in predicting stock market in emerging and mature economies. To the best of our knowledge, there have been no prior studies on the use of MARSplines to forecast high frequency data in mature and emerging markets. This paper provides new useful information to intraday traders about the use of these techniques in forecasting stock index or stocks in emerging economies. Our study shows that MARSPlines and SVR perform better than BPNN in predicting the stock market index values, the former being the best of the three for the prediction of NCI and the latter being best suited for Nifty.