Data mining techniques are gaining an important place in finance as the size of the data is increasing exponentially with every passing day and the accuracy with which the data should be analyzed is very important. In this paper, an attempt is made to develop two models, one using three-layer feed-forward back propagation Artificial Neural Network (ANN 4-4-1) and the other using Autoregressive Integrated Moving Average (ARIMA 1, 1, 1) for forecasting the future index value of Sensex (BSE 30). Simulations have been done using prices of daily open, high, low and close of Sensex. These are chosen as input data values and output is the forecasted closing price of Sensex for next day. Convergence and performance of models have been evaluated on the basis of the simulation results.

Applications of artificial neural networks to financial forecasting have become very popular over the last few years [see, for instance, Gately (1996), Yao et al. (1999), Abu-Mostafa et al. (2001) and Zhang, et al. (2004)]. It is a natural way for solving problems that involve learning and pattern recognition. It can detect patterns in data through learning; they are much easier to program since they elicit general rules from examples (Afolabi and Olude, 2007). Neural networks can approximate unknown functions to any degree of desired accuracy without making any unnecessary assumptions about the distribution of the data (Sexton and Sikander, 2001). Neural networks can make it possible to approximate both linear and non-linear functions and achieve outstanding performance. Neural networks have become popular in the world of forecasting because of their non-parametric approach (Dacha, 2007).

In sharp contrast to this, the traditional approaches to time series such as ARIMA assume that the series under study is generated from linear progresses and linear models have an advantage in that they can be understood and analyzed in greater detail (Box and Jenkins, 1976). The ARIMA model is basically used for non-stationary series when linearity between variables is assumed (Gomes et al., 2006). It assumes that future value of time series is linearly related to the past observations.

In this paper, Artificial Neural Network (ANN), using feed-forward back propagation supervised learning, and Autoregressive Integrated Moving Average (ARIMA) are developed and compared. They are then used for forecasting the next day's close value of Sensex. This paper then goes on to evaluate the performance of these two models by calculating four types of errors.

Applied Finance Journal, Next Day Stock Market Forecasting, Data Mining Techniques, Artificial Neural Networks, Financial Markets, Efficient Market Hypothesis, Multilayer Neural Network, Network Architecture, Mean Square Error, Transformation Process, Bayesian Information Criteria.